HBase默认配置

官网配置文档：http://hbase.apache.org/book.html#_configuration_files
中文翻译转自：http://eclecl1314-163-com.iteye.com/blog/1474286

■ 常见重要配置参数

(1) hbase.rpc.timeout
rpc的超时时间，默认 60s，不建议修改，避免影响正常的业务，在线上环境刚开始配置的是 3 秒，运行半天后发现了大量的 timeout error，原因是有一个 region 出现了如下问题阻塞了写操作：“Blocking updates … memstore size 434.3m is >= than blocking 256.0m size”可见不能太低
(2) ipc.socket.timeout：socket 建立链接的超时时间，应该小于或者等于 rpc 的超时时间，默认为 20s
(3) hbase.client.retries.number：重试次数，默认为 10，可配置为 3
(4) hbase.client.pause：重试的休眠时间，默认为 1s，可减少，比如 100ms
(5) hbase.regionserver.lease.period：scan 查询时每次与 server 交互的超时时间，默认为 60s，可不调整

■ HBase默认配置参数

1. hbase.rootdir

  • 这个目录是 RegionServer 的共享目录，用来持久化 HBase。URL 需要是 “完全正确” 的，还要包含文件系统的 scheme。 例如 “/hbase” 表示 HBase 在 HDFS 中占用的实际存储位置，HDFS 的 NameNode 运行在主机名为 master5 的 8020 端口，则 hbase.rootdir 的设置应为 “hdfs://master5:8020/hbase”。在默认情况下 HBase 是写在 /tmp 中的。不修改这个配置的话，数据会在重启的时候丢失。特别注意的是 hbase.rootdir 里面的 HDFS 地址是要跟 Hadoop 的 core-site.xml 里面的 fs.defaultFS 的 HDFS 的 IP 地址或者域名、端口必须一致。
  • 默认为 file:///tmp/hbase-${user.name}/hbase

2. hbase.cluster.distributed

  • HBase 的运行模式。为 false 表示单机模式，为 true 表示分布式模式。若为 false，HBase 和 ZooKeeper 会运行在同一个 JVM 中
  • 默认值为 false

3. hbase.master （hbase.master.port）

  • 如果只设置单个 Hmaster，那么 hbase.master 属性参数需要设置为 master5:60000 (主机名:60000)
  • 如果要设置多个 Hmaster，那么我们只需要提供端口 60000，因为选择真正的 master 的事情会有 zookeeper 去处理
  • 默认端口号：60000

4. hbase.tmp.dir

  • 本地文件系统的临时文件夹。可以修改到一个更为持久的目录上。(/tmp会在重启时清除)
  • 默认: /tmp/hbase-${user.name}

5. hbase.master.info.port

  • HBase Master web 界面端口. 设置为 -1 意味着你不想让它运行
  • 默认: 60010

6. hbase.master.info.bindAddress

  • HBase Master web 界面绑定的端口
  • 默认: 0.0.0.0

7. hbase.client.write.buffer

  • HTable 客户端的写缓冲的默认大小。这个值越大，需要消耗的内存越大。因为缓冲在客户端和服务端都有实例，所以需要消耗客户端和服务端两个地方的内存。得到的好处是，可以减少 RPC 的次数。可以这样估算服务器端被占用的内存： hbase.client.write.buffer * hbase.regionserver.handler.count
  • 默认: 2097152 (2G)

8. hbase.regionserver.port

  • HBase RegionServer 绑定的端口
  • 默认: 60020

9. hbase.regionserver.info.port

  • HBase RegionServer web 界面绑定的端口 设置为 -1 意味这你不想与运行 RegionServer 界面
  • 默认: 60030

10. hbase.regionserver.info.port.auto

  • Master 或 RegionServer 是否要动态搜一个可以用的端口来绑定界面。当 hbase.regionserver.info.port已经被占用的时候，可以搜一个空闲的端口绑定。这个功能在测试的时候很有用。默认关闭
  • 默认: false

11. hbase.regionserver.info.bindAddress

  • HBase RegionServer web 界面的IP地址
  • 默认: 0.0.0.0

12. hbase.regionserver.class

  • RegionServer 使用的接口。客户端打开代理来连接region server的时候会使用到
  • 默认: org.apache.hadoop.hbase.ipc.HRegionInterface

13. hbase.client.pause

  • 通常的客户端暂停时间。最多的用法是客户端在重试前的等待时间。比如失败的 get 操作和 region 查询操作等都很可能用到
  • 默认: 1000 ms

14. hbase.client.retries.number

  • 最大重试次数。例如 region 查询，Get 操作，Update 操作等等都可能发生错误，需要重试。这是最大重试错误的值
  • 默认: 10

15. hbase.client.scanner.caching

  • 当调用 Scanner 的 next 方法，而值又不在缓存里的时候，从服务端一次获取的行数。越大的值意味着 Scanner 会快一些，但是会占用更多的内存。当缓冲被占满的时候，next 方法调用会越来越慢。慢到一定程度，可能会导致超时。例如超过了 hbase.regionserver.lease.period
  • 默认: 1

16.hbase.client.keyvalue.maxsize

  • 一个 KeyValue 实例的最大 size。这个是用来设置存储文件中的单个 entry 的大小上界。因为一个 KeyValue 是不能分割的，所以可以避免因为数据过大导致 region 不可分割。明智的做法是把它设为可以被最大 region size 整除的数。如果设置为 0 或者更小，就会禁用这个检查。默认 10MB
  • 默认: 10485760 (10MB)

17. hbase.regionserver.lease.period

  • 客户端租用 HRegion server 期限，即超时阀值。单位是毫秒。默认情况下，客户端必须在这个时间内发一条信息，否则视为死掉。
  • 默认: 60000

18. hbase.regionserver.handler.count

  • RegionServers 受理的 RPC Server 实例数量。对于 Master 来说，这个属性是 Master 受理的 handler 数量
  • 默认: 10

19. hbase.regionserver.msginterval

  • RegionServer 发消息给 Master 时间间隔，单位是毫秒
  • 默认: 3000

20. hbase.regionserver.optionallogflushinterval

  • 将 Hlog 同步到 HDFS 的间隔。如果 Hlog 没有积累到一定的数量，到了时间，也会触发同步。默认是 1秒，单位毫秒。
  • 默认: 1000

21. hbase.regionserver.regionSplitLimit

  • region 的数量到了这个值后就不会在分裂了。这不是一个 region 数量的硬性限制。但是起到了一定指导性的作用，到了这个值就该停止分裂了。默认是 MAX_INT。就是说不阻止分裂。
  • 默认: 2147483647 (1G)

22. hbase.regionserver.logroll.period

  • 提交 commit log 的间隔，不管有没有写足够的值
  • 默认: 3600000

23. hbase.regionserver.hlog.reader.impl

  • HLog file reader 的实现
  • 默认: org.apache.hadoop.hbase.regionserver.wal.SequenceFileLogReader

24. hbase.regionserver.hlog.writer.impl

  • HLog file writer 的实现
  • 默认: org.apache.hadoop.hbase.regionserver.wal.SequenceFileLogWriter

25. hbase.regionserver.thread.splitcompactcheckfrequency

  • region server 多久执行一次 split/compaction 检查
  • 默认: 20000

26. hbase.regionserver.nbreservationblocks

  • 储备的内存 block 的数量(译者注:就像石油储备一样)。当发生 out of memory 异常的时候，我们可以用这些内存在 RegionServer 停止之前做清理操作
  • 默认: 4

27. hbase.zookeeper.dns.interface

  • 当使用 DNS 的时候，Zookeeper 用来上报的 IP 地址的网络接口名字
  • 默认: default

28. hbase.zookeeper.dns.nameserver

  • 当使用 DNS 的时候，Zookeepr 使用的 DNS 的域名或者 IP 地址，Zookeeper 用它来确定和 master 用来进行通讯的域名
  • 默认: default

29. hbase.regionserver.dns.interface

  • 当使用 DNS 的时候，RegionServer 用来上报的 IP 地址的网络接口名字
  • 默认: default

30. hbase.regionserver.dns.nameserver

  • 当使用 DNS 的时候，RegionServer 使用的 DNS 的域名或者 IP 地址，RegionServer 用它来确定和 master 用来进行通讯的域名
  • 默认: default

31. hbase.master.dns.interface

  • 当使用 DNS 的时候，Master 用来上报的 IP 地址的网络接口名字
  • 默认: default

32. hbase.master.dns.nameserver

  • 当使用DNS的时候，RegionServer使用的DNS的域名或者IP地址，Master用它来确定用来进行通讯的域名
  • 默认: default

33. hbase.balancer.period

  • Master 执行 region balancer 的间隔
  • 默认: 300000

34. hbase.regions.slop

  • 当任一 regionserver 有 average + (average * slop) 个 region 是会执行 Rebalance
  • 默认: 0

35. hbase.master.logcleaner.ttl

  • Hlog 存在于.oldlogdir 文件夹的最长时间, 超过了就会被 Master 的线程清理掉
  • 默认: 600000

36. hbase.master.logcleaner.plugins

  • LogsCleaner 服务会执行的一组 LogCleanerDelegat。值用逗号间隔的文本表示。这些 WAL/HLog cleaners 会按顺序调用。可以把先调用的放在前面。你可以实现自己的 LogCleanerDelegat，加到 Classpath下，然后在这里写 下类的全称。一般都是加在默认值的前面
  • 默认: org.apache.hadoop.hbase.master.TimeToLiveLogCleaner

37. hbase.regionserver.global.memstore.upperLimit

  • 单个 region server 的全部 memtores 的最大值。超过这个值，一个新的 update 操作会被挂起，强制执行flush操作
  • 默认: 0.4

38. hbase.regionserver.global.memstore.lowerLimit

  • 当强制执行 flush 操作的时候，当低于这个值的时候，flush 会停止。默认是堆大小的 35% . 如果这个值和 hbase.regionserver.global.memstore.upperLimit 相同就意味着当 update 操作因为内存限制被挂起时，会尽量少的执行 flush(译者注:一旦执行 flush，值就会比下限要低，不再执行)
  • 默认: 0.35

39. hbase.server.thread.wakefrequency

  • service 工作的 sleep 间隔，单位毫秒。 可以作为 service 线程的 sleep 间隔，比如 log roller
  • 默认: 10000

40. hbase.hregion.memstore.flush.size

  • 当 memstore 的大小超过这个值的时候，会 flush 到磁盘。这个值被一个线程每隔 hbase.server.thread.wakefrequency 检查一下
  • 默认: 67108864 (64MB)

41. hbase.hregion.preclose.flush.size

  • 当一个 region 中的 memstore 的大小大于这个值的时候，我们又触发了 close. 会先运行“pre-flush”操作，清理这个需要关闭的 memstore，然后将这个 region 下线。当一个 region 下线了，我们无法再进行任何写操作。如果一个 memstore 很大的时候，flush 操作会消耗很多时间。”pre-flush” 操作意味着在 region 下线之前，会先把 memstore 清空。这样在最终执行 close 操作的时候，flush 操作会很快
  • 默认: 5242880 (5MB)

42. hbase.hregion.memstore.block.multiplier

  • 如果 memstore 有 hbase.hregion.memstore.block.multiplier 倍数的 hbase.hregion.flush.size 的大小，就会阻塞 update 操作。这是为了预防在 update 高峰期会导致的失控。如果不设上 界，flush 的时候会花很长的时间来合并或者分割，最坏的情况就是引发 out of memory 异常。(译者注:内存操作的速度和磁盘不匹配，需要等一等。原文似乎有误)
  • 默认: 2

43. hbase.hregion.memstore.mslab.enabled

  • 体验特性：启用 memStore 分配本地缓冲区。这个特性是为了防止在大量写负载的时候堆的碎片过多。这可以减少 GC 操作的频率。( GC 有可能会 Stop the world )(译者注：实现的原理相当于预分配内存，而不是每一个值都要从堆里分配)
  • 默认: false

44. hbase.hregion.max.filesize

  • 最大 HStoreFile 大小。若某个 Column families 的 HStoreFile 增长达到这个值，这个 Hegion 会被切割成两个
  • 默认: 268435456 (256M)

45. hbase.hstore.compactionThreshold

  • 当一个 HStore 含有多于这个值的 HStoreFiles (每一个 memstore flush 产生一个 HStoreFile )的时候，会执行一个合并操作，把这 HStoreFiles 写成一个。这个值越大，需要合并的时间就越长
  • 默认: 3

46. hbase.hstore.blockingStoreFiles

  • 当一个 HStore 含有多于这个值的 HStoreFiles (每一个 memstore flush 产生一个 HStoreFile )的时候，会执行一个合并操作，update 会阻塞直到合并完成，直到超过了 hbase.hstore.blockingWaitTime 的值
  • 默认: 7

47. hbase.hstore.blockingWaitTime

  • hbase.hstore.blockingStoreFiles 所限制的 StoreFile 数量会导致 update 阻塞，这个时间是来限制阻塞时间的。当超过了这个时间，HRegion 会停止阻塞update操作，不过合并还有没有完成。默认为 90s
  • 默认: 90000

48. hbase.hstore.compaction.max

  • 每个“小”合并的 HStoreFiles 最大数量
  • 默认: 10

49. hbase.hregion.majorcompaction

  • 一个 Region 中的所有 HStoreFile 的 major compactions 的时间间隔。默认是 1 天。 设置为 0 就是禁用这个功能
  • 默认: 86400000

50. hbase.mapreduce.hfileoutputformat.blocksize

  • MapReduce 中 HFileOutputFormat 可以写 storefiles/hfiles。这个值是 hfile 的 blocksize 的最小值。通常在 Hbase写Hfile 的时候，bloocksize 是由 table schema(HColumnDescriptor) 决定的，但是在 mapreduce 写的时候，我们无法获取 schema 中 blocksize。这个值 越小，你的索引就越大，你随机访问需要获取的数据就越小。如果你的 cell 都很小，而且你需要更快的随机访问，可以把这个值调低
  • 默认: 65536

51. hfile.block.cache.size

  • 分配给 HFile/StoreFile 的 block cache 占最大堆(-Xmx setting)的比例。默认是20%，设置为0就是不分配
  • 默认: 0.2

52. hbase.hash.type

  • 哈希函数使用的哈希算法。可以选择两个值:: murmur (MurmurHash) 和 jenkins (JenkinsHash). 这个哈希是给 bloom filters 用的
  • 默认: murmur

53. hbase.master.keytab.file

  • HMaster server 验证登录使用的 kerberos keytab 文件路径。(译者注：Hbase 使用 Kerberos实现安全)
  • 默认: none

54. hbase.master.kerberos.principal

  • 例如. “hbase/_HOST@EXAMPLE.COM”。 HMaster 运行需要使用 kerberos principal name. principal name 可以在: user/hostname@DOMAIN 中获取. 如果 “_HOST” 被用做 hostname portion，需要使用实际运行的 hostname 来替代它
  • 默认: none

55. hbase.regionserver.keytab.file

  • HRegionServer 验证登录使用的 kerberos keytab 文件路径
  • 默认: none

56. hbase.regionserver.kerberos.principal

  • 例如. “hbase/_HOST@EXAMPLE.COM”。HRegionServer 运行需要使用 kerberos principal name. principal name 可以在: user/hostname@DOMAIN 中获取。如果 “_HOST” 被用做 hostname portion，需要使用实际运行的 hostname 来替代它。在这个文件中必须要有一个entry来描述 hbase.regionserver.keytab.file
  • 默认: none

57. zookeeper.session.timeout

  • ZooKeeper 会话超时。Hbase 把这个值传递改 zk 集群，向它推荐一个会话的最大超时时间。单位是毫秒
  • 默认: 180000

58. zookeeper.znode.parent

  • ZooKeeper 中的 Hbase 的根 ZNode。所有的 Hbase 的 ZooKeeper 会用这个目录配置相对路径。默认情况下，所有的 Hbase 的 ZooKeeper 文件路径是用相对路径，所以它们会都去这个目录下面\
  • 默认: /hbase

59. zookeeper.znode.rootserver

  • ZNode 保存的根 region 的路径. 这个值是由 Master 来写，client 和 regionserver 来读的。如果设为一个相对地址，父目录就是 ${zookeeper.znode.parent}。默认情形下，意味着根 region的路径存储在/hbase/root-region- server
  • 默认: root-region-server

60. hbase.zookeeper.quorum

  • Zookeeper 集群的地址列表，用逗号分割。例如：”host1.mydomain.com,host2.mydomain.com,host3.mydomain.com”.默认是 localhost,是给伪分布式用的。要修改才能在完全分布式的情况下使用。如果在hbase-env.sh设置了HBASE_MANAGES_ZK， 这些ZooKeeper节点就会和Hbase一起启动
  • 默认: localhost

61. hbase.zookeeper.peerport

  • ZooKeeper节点使用的端口
  • 默认: 2888

62. hbase.zookeeper.leaderport

  • ZooKeeper用来选择Leader的端口
  • 默认: 3888

63. hbase.zookeeper.property.initLimit

  • ZooKeeper的zoo.conf中的配置。 初始化synchronization阶段的ticks数量限制
  • 默认: 10

64. hbase.zookeeper.property.syncLimit

  • ZooKeeper的zoo.conf中的配置。 发送一个请求到获得承认之间的ticks的数量限制
  • 默认: 5

65. hbase.zookeeper.property.dataDir

  • ZooKeeper的zoo.conf中的配置。 快照的存储位置
  • 默认: ${hbase.tmp.dir}/zookeeper

66. hbase.zookeeper.property.clientPort

  • ZooKeeper的zoo.conf中的配置。 客户端连接的端口
  • 默认: 2181

67. hbase.zookeeper.property.maxClientCnxns

  • ZooKeeper的zoo.conf中的配置。 ZooKeeper集群中的单个节点接受的单个Client(以IP区分)的请求的并发数。这个值可以调高一点，防止在单机和伪分布式模式中出问题。
  • 默认: 2000

68. hbase.rest.port

  • HBase REST server的端口
  • 默认: 8080

69. hbase.rest.readonly

  • 定义REST server的运行模式。可以设置成如下的值： false: 所有的HTTP请求都是被允许的 - GET/PUT/POST/DELETE. true:只有GET请求是被允许的
  • 默认: false

70. hbase.regionserver.restart.on.zk.expire

  • 当 regionserver 遇到 ZooKeeper session expired ， regionserver 是否选择 restart
  • 默认: false

    HBase配置常量类

    org.apache.hadoop.hbase.HConstants
key 含义 默认值
hbase.zookeeper.recoverable.waittime zookeeper恢复的等待时间 10000
hbase.zookeeper.property.maxClientCnxns zookeeper并发连接限制 300
zookeeper.session.timeout zookeeper会话超时时间 180 * 1000
hbase.zookeeper.useMulti 是否使用ZK的multi-update操作（多个原子操作合并，保证结果的一致性） false
hbase.regionserver.port region server监听端口 60020
hbase.regionserver.info.port 默认的regionserverin信息端口 60030
hbase.server.thread.wakefrequency 线程唤醒频率 10 * 1000
hbase.server.versionfile.writeattempts 多久写入版本文件，失败前 3
hbase.hstore.compaction.kv.max 批量flush/compaction的最大KV数 10
hbase.client.ipc.pool.type Hbase客户端IPC Pool类型 PoolType.RoundRobin
hbase.client.ipc.pool.size Hbase客户端IPC Pool大小 1
hbase.client.operation.timeout Hbase客户端操作超时时间（覆盖RPC超时时间） Integer.MAX_VALUE
hbase.client.meta.operation.timeout Hbase客户端操作超时时间（覆盖RPC超时时间） Integer.MAX_VALUE
hbase.hregion.max.filesize 切分的region最大文件大小 10 1024 1024 * 1024
hbase.hstore.open.and.close.threads.max 打开/关闭存储或并行存储的线程数 1
hbase.hregion.edits.replay.skip.errors 跳过错误重编辑 false
hbase.client.scanner.max.result.size scan最大的字节数 Long.MAX_VALUE
hbase.client.pause 一次get失败或region搜寻之后客户端暂停时间（失败睡眠时间） 100
hbase.client.max.total.tasks 客户端维护的最大并发连接 100
hbase.client.max.perserver.tasks 客户端对于一个regionserver维护的最大并发连接 2
hbase.client.max.perregion.tasks 客户端对于一个region维护的最大并发连接 1
hbase.server.pause 失败操作后的重试等待时间 1000
hbase.client.retries.number 客户端重试次数 31
hbase.client.prefetch.limit 预取的region个数限制 10
hbase.client.scanner.caching 所有客户端的默认scan缓存行数rows 100
hbase.meta.scanner.caching 元数据表（hbase:meta）的scan缓存行数rows 100
hbase.client.scanner.timeout.period 客户端scan超时时间 60000 ms
hbase.rpc.timeout hbase rpc超时时间 60000
hbase.rpc.shortoperation.timeout rpc短操作超时时间 10000
hbase.client.write.buffer 客户端写入数据缓冲区 2097152≈2M
hbase.client.keyvalue.maxsize 客户端最长的keyvalue值 -1（不限制）
hbase.ipc.client.connection.maxidletime 客户端连接最大空闲时间 10000（10s）
hbase.ipc.client.connect.max.retries 客户端连接最大重试次数 0
hbase.ipc.client.tcpnodelay tcp无延迟 true
hbase.ipc.client.tcpkeepalive tcp保活 true
ipc.ping.interval 客户端ping频率 60000（1 min）
ipc.socket.timeout 发起连接超时时间 20000（20s）

HBase配置文件

hbase-default.xml

  1. <?xml version="1.0"?>  
  2. <?xml-stylesheet type="text/xsl" href="configuration.xsl"?>  
  3. <configuration>  
  4.     <!-- hbase的本地临时目录,每次机器重启数据会丢失，建议放到某个持久化文件目录下 -->  
  5.     <property>  
  6.         <name>hbase.tmp.dir</name>  
  7.         <value>${java.io.tmpdir}/hbase-${user.name}</value>  
  8.         <description>Temporary directory on the local filesystem.  
  9.             Change this setting to point to a location more permanent  
  10.             than '/tmp', the usual resolve for java.io.tmpdir, as the  
  11.             '/tmp' directory is cleared on machine restart.  
  12.         </description>  
  13.     </property>  
  14.     <!-- 每个regionServer的共享目录,用来持久化Hbase,默认情况下在/tmp/hbase下面 -->  
  15.     <property>  
  16.         <name>hbase.rootdir</name>  
  17.         <value>${hbase.tmp.dir}/hbase</value>  
  18.         <description>The directory shared by region servers and into  
  19.             which HBase persists. The URL should be 'fully-qualified'  
  20.             to include the filesystem scheme. For example, to specify the  
  21.             HDFS directory '/hbase' where the HDFS instance's namenode is  
  22.             running at namenode.example.org on port 9000, set this value to:  
  23.             hdfs://namenode.example.org:9000/hbase. By default, we write  
  24.             to whatever ${hbase.tmp.dir} is set too -- usually /tmp --  
  25.             so change this configuration or else all data will be lost on  
  26.             machine restart.  
  27.         </description>  
  28.     </property>  
  29.     <!-- hbase底层如果使用hdfs作为文件系统，这里是指默认在文件系统的临时存储目录用来存储临时数据 -->  
  30.     <property>  
  31.         <name>hbase.fs.tmp.dir</name>  
  32.         <value>/user/${user.name}/hbase-staging</value>  
  33.         <description>A staging directory in default file system (HDFS)  
  34.             for keeping temporary data.  
  35.         </description>  
  36.     </property>  
  37.     <!-- hdfs里面批量装载的目录 -->  
  38.     <property>  
  39.         <name>hbase.bulkload.staging.dir</name>  
  40.         <value>${hbase.fs.tmp.dir}</value>  
  41.         <description>A staging directory in default file system (HDFS)  
  42.             for bulk loading.  
  43.         </description>  
  44.     </property>  
  45.     <!-- hbase集群模式,false表示hbase的单机，true表示是分布式模式 -->  
  46.     <property>  
  47.         <name>hbase.cluster.distributed</name>  
  48.         <value>false</value>  
  49.         <description>The mode the cluster will be in. Possible values are  
  50.             false for standalone mode and true for distributed mode. If  
  51.             false, startup will run all HBase and ZooKeeper daemons together  
  52.             in the one JVM.  
  53.         </description>  
  54.     </property>  
  55.     <!-- hbase依赖的zk地址 -->  
  56.     <property>  
  57.         <name>hbase.zookeeper.quorum</name>  
  58.         <value>localhost</value>  
  59.         <description>Comma separated list of servers in the ZooKeeper ensemble  
  60.             (This config. should have been named hbase.zookeeper.ensemble).  
  61.             For example, "host1.mydomain.com,host2.mydomain.com,host3.mydomain.com".  
  62.             By default this is set to localhost for local and pseudo-distributed  
  63.             modes  
  64.             of operation. For a fully-distributed setup, this should be set to a  
  65.             full  
  66.             list of ZooKeeper ensemble servers. If HBASE_MANAGES_ZK is set in  
  67.             hbase-env.sh  
  68.             this is the list of servers which hbase will start/stop ZooKeeper on as  
  69.             part of cluster start/stop. Client-side, we will take this list of  
  70.             ensemble members and put it together with the  
  71.             hbase.zookeeper.clientPort  
  72.             config. and pass it into zookeeper constructor as the connectString  
  73.             parameter.  
  74.         </description>  
  75.     </property>  
  76.     <!-- 如果是本地存储，位于本地文件系统的路径 -->  
  77.     <property>  
  78.         <name>hbase.local.dir</name>  
  79.         <value>${hbase.tmp.dir}/local/</value>  
  80.         <description>Directory on the local filesystem to be used  
  81.             as a local storage.  
  82.         </description>  
  83.     </property>  
  85.     <!-- hbase master节点的端口 -->  
  86.     <property>  
  87.         <name>hbase.master.port</name>  
  88.         <value>16000</value>  
  89.         <description>The port the HBase Master should bind to.</description>  
  90.     </property>  
  91.     <!-- hbase master的web ui页面的端口 -->  
  92.     <property>  
  93.         <name>hbase.master.info.port</name>  
  94.         <value>16010</value>  
  95.         <description>The port for the HBase Master web UI.  
  96.             Set to -1 if you do not want a UI instance run.  
  97.         </description>  
  98.     </property>  
  99.     <!-- hbase master的web ui页面绑定的地址 -->  
  100.     <property>  
  101.         <name>hbase.master.info.bindAddress</name>  
  102.         <value>0.0.0.0</value>  
  103.         <description>The bind address for the HBase Master web UI  
  104.         </description>  
  105.     </property>  
  106.     <!-- 不知道干什么用 -->  
  107.     <property>  
  108.         <name>hbase.master.logcleaner.plugins</name>  
  109.         <value>org.apache.hadoop.hbase.master.cleaner.TimeToLiveLogCleaner  
  110.         </value>  
  111.         <description>A comma-separated list of BaseLogCleanerDelegate invoked  
  112.             by  
  113.             the LogsCleaner service. These WAL cleaners are called in order,  
  114.             so put the cleaner that prunes the most files in front. To  
  115.             implement your own BaseLogCleanerDelegate, just put it in HBase's classpath  
  116.             and add the fully qualified class name here. Always add the above  
  117.             default log cleaners in the list.  
  118.         </description>  
  119.     </property>  
  120.     <!-- hbase清理oldlogdir目录下的hlog文件的最长时间 ，单位毫秒 -->  
  121.     <property>  
  122.         <name>hbase.master.logcleaner.ttl</name>
  123.         <value>600000</value>  
  124.         <description>Maximum time a WAL can stay in the .oldlogdir directory,  
  125.             after which it will be cleaned by a Master thread.  
  126.         </description>  
  127.     </property>  
  128.     <property>  
  129.         <name>hbase.master.hfilecleaner.plugins</name>  
  130.         <value>org.apache.hadoop.hbase.master.cleaner.TimeToLiveHFileCleaner  
  131.         </value>  
  132.         <description>A comma-separated list of BaseHFileCleanerDelegate  
  133.             invoked by  
  134.             the HFileCleaner service. These HFiles cleaners are called in order,  
  135.             so put the cleaner that prunes the most files in front. To  
  136.             implement your own BaseHFileCleanerDelegate, just put it in HBase's classpath  
  137.             and add the fully qualified class name here. Always add the above  
  138.             default log cleaners in the list as they will be overwritten in  
  139.             hbase-site.xml.  
  140.         </description>  
  141.     </property>  
  142.     <!-- Catalog Janitor从master到META的超时时间，我们知道这个Janitor是定时的去META扫描表目录，来决定回收无用的regions -->  
  143.     <property>  
  144.         <name>hbase.master.catalog.timeout</name>  
  145.         <value>600000</value>  
  146.         <description>Timeout value for the Catalog Janitor from the master to META.  
  147.         </description>  
  148.     </property>  
  149.     <!-- master是否监听master web ui端口并重定向请求给web ui服务器，该配置是master和RegionServer共享 -->  
  150.     <property>  
  151.         <name>hbase.master.infoserver.redirect</name>  
  152.         <value>true</value>  
  153.         <description>Whether or not the Master listens to the Master web  
  154.             UI port (hbase.master.info.port) and redirects requests to the web  
  155.             UI server shared by the Master and RegionServer.  
  156.         </description>  
  157.     </property>  
  159.     <!-- hbase regionServer的默认端口 -->  
  160.     <property>  
  161.         <name>hbase.regionserver.port</name>  
  162.         <value>16020</value>  
  163.         <description>The port the HBase RegionServer binds to.</description>  
  164.     </property>  
  165.     <!-- hbase regionServer的web ui的默认端口 -->  
  166.     <property>  
  167.         <name>hbase.regionserver.info.port</name>  
  168.         <value>16030</value>  
  169.         <description>The port for the HBase RegionServer web UI  
  170.             Set to -1 if you do not want the RegionServer UI to run.  
  171.         </description>  
  172.     </property>  
  173.     <!-- hbase regionServer的web ui绑定地址 -->  
  174.     <property>  
  175.         <name>hbase.regionserver.info.bindAddress</name>  
  176.         <value>0.0.0.0</value>  
  177.         <description>The address for the HBase RegionServer web UI  
  178.         </description>  
  179.     </property>  
  180.     <!-- 如果regionServer默认的端口被暂用了，是否允许hbase搜索一个可用的端口来绑定 -->  
  181.     <property>  
  182.         <name>hbase.regionserver.info.port.auto</name>  
  183.         <value>false</value>  
  184.         <description>Whether or not the Master or RegionServer  
  185.             UI should search for a port to bind to. Enables automatic port  
  186.             search if hbase.regionserver.info.port is already in use.  
  187.             Useful for testing, turned off by default.  
  188.         </description>  
  189.     </property>  
  190.     <!-- regionServer端默认开启的RPC监控实例数，也即RegionServer能够处理的IO请求线程数 -->  
  191.     <property>  
  192.         <name>hbase.regionserver.handler.count</name>  
  193.         <value>30</value>  
  194.         <description>Count of RPC Listener instances spun up on RegionServers.  
  195.             Same property is used by the Master for count of master handlers.  
  196.         </description>  
  197.     </property>  
  198.     <!-- hbase提供的可以用来处理请求的队列数 0.1 * 总数，如果为0则表示所有请求公用一个队列， 如果为1则表示每个请求自己有一个独立的队列 -->  
  199.     <property>  
  200.         <name>hbase.ipc.server.callqueue.handler.factor</name>  
  201.         <value>0.1</value>  
  202.         <description>Factor to determine the number of call queues.  
  203.             A value of 0 means a single queue shared between all the handlers.  
  204.             A value of 1 means that each handler has its own queue.  
  205.         </description>  
  206.     </property>  
  207.     <!-- hbase提供的读写队列数比例,参数值为0-1之间，如果为0则所有队列同时处理读写请求 -->  
  208.     <!-- 现在假设我们有10个队列 1、该值设置为0，则这10个队列同时处理读写请求 2、该值设置为1，则1个队列处理写情况，9个队列处理读请求   
  209.         3、该值设置为0.x,则x个队列处理处理读请求，10-x个队列处理写请求 4、根据实际情况，读多写少还是写少读多，可按需配置 -->  
  210.     <property>  
  211.         <name>hbase.ipc.server.callqueue.read.ratio</name>  
  212.         <value>0</value>  
  213.         <description>Split the call queues into read and write queues.  
  214.             The specified interval (which should be between 0.0 and 1.0)  
  215.             will be multiplied by the number of call queues.  
  216.             A value of 0 indicate to not split the call queues, meaning that both  
  217.             read and write  
  218.             requests will be pushed to the same set of queues.  
  219.             A value lower than 0.5 means that there will be less read queues than  
  220.             write queues.  
  221.             A value of 0.5 means there will be the same number of read and write  
  222.             queues.  
  223.             A value greater than 0.5 means that there will be more read queues  
  224.             than write queues.  
  225.             A value of 1.0 means that all the queues except one are used to  
  226.             dispatch read requests.  
  228.             Example: Given the total number of call queues being 10  
  229.             a read.ratio of 0 means that: the 10 queues will contain both  
  230.             read/write requests.  
  231.             a read.ratio of 0.3 means that: 3 queues will contain only read  
  232.             requests  
  233.             and 7 queues will contain only write requests.  
  234.             a read.ratio of 0.5 means that: 5 queues will contain only read  
  235.             requests  
  236.             and 5 queues will contain only write requests.  
  237.             a read.ratio of 0.8 means that: 8 queues will contain only read  
  238.             requests  
  239.             and 2 queues will contain only write requests.  
  240.             a read.ratio of 1 means that: 9 queues will contain only read requests  
  241.             and 1 queues will contain only write requests.  
  242.         </description>  
  243.     </property>  
  244.     <!-- hbase提供的用于支持get/scan请求的队列比例 -->  
  245.     <property>  
  246.         <name>hbase.ipc.server.callqueue.scan.ratio</name>  
  247.         <value>0</value>  
  248.         <description>Given the number of read call queues, calculated from the  
  249.             total number  
  250.             of call queues multiplied by the callqueue.read.ratio, the scan.ratio  
  251.             property  
  252.             will split the read call queues into small-read and long-read queues.  
  253.             A value lower than 0.5 means that there will be less long-read queues  
  254.             than short-read queues.  
  255.             A value of 0.5 means that there will be the same number of short-read  
  256.             and long-read queues.  
  257.             A value greater than 0.5 means that there will be more long-read  
  258.             queues than short-read queues  
  259.             A value of 0 or 1 indicate to use the same set of queues for gets and  
  260.             scans.  
  262.             Example: Given the total number of read call queues being 8  
  263.             a scan.ratio of 0 or 1 means that: 8 queues will contain both long and  
  264.             short read requests.  
  265.             a scan.ratio of 0.3 means that: 2 queues will contain only long-read  
  266.             requests  
  267.             and 6 queues will contain only short-read requests.  
  268.             a scan.ratio of 0.5 means that: 4 queues will contain only long-read  
  269.             requests  
  270.             and 4 queues will contain only short-read requests.  
  271.             a scan.ratio of 0.8 means that: 6 queues will contain only long-read  
  272.             requests  
  273.             and 2 queues will contain only short-read requests.  
  274.         </description>  
  275.     </property>  
  276.     <!-- regionServer发送消息给Master的时间间隔,单位是毫秒 -->  
  277.     <property>  
  278.         <name>hbase.regionserver.msginterval</name>  
  279.         <value>3000</value>  
  280.         <description>Interval between messages from the RegionServer to Master  
  281.             in milliseconds.  
  282.         </description>  
  283.     </property>  
  284.     <!-- regionServer日志滚动提交的周期,不管这个日志有没有写满 -->  
  285.     <property>  
  286.         <name>hbase.regionserver.logroll.period</name>  
  287.         <value>3600000</value>  
  288.         <description>Period at which we will roll the commit log regardless  
  289.             of how many edits it has.  
  290.         </description>  
  291.     </property>  
  292.     <!-- 在regionServer上的WAL日志，在停止服务前允许的关闭 WAL 的连续错误数量 比如如果我们日志在滚动提交的是，此时wal写入错误，那么就会立即停止regionServer的服务   
  293.         默认值2表示运行有2个错误发生 -->  
  294.     <property>  
  295.         <name>hbase.regionserver.logroll.errors.tolerated</name>  
  296.         <value>2</value>  
  297.         <description>The number of consecutive WAL close errors we will allow  
  298.             before triggering a server abort. A setting of 0 will cause the  
  299.             region server to abort if closing the current WAL writer fails during  
  300.             log rolling. Even a small value (2 or 3) will allow a region server  
  301.             to ride over transient HDFS errors.  
  302.         </description>  
  303.     </property>  
  304.     <!-- regionServer的WAL文件读取的实现类 -->  
  305.     <property>  
  306.         <name>hbase.regionserver.hlog.reader.impl</name>  
  307.         <value>org.apache.hadoop.hbase.regionserver.wal.ProtobufLogReader  
  308.         </value>  
  309.         <description>The WAL file reader implementation.</description>  
  310.     </property>  
  311.     <!-- regionServer的WAL文件写的实现类 -->  
  312.     <property>  
  313.         <name>hbase.regionserver.hlog.writer.impl</name>  
  314.         <value>org.apache.hadoop.hbase.regionserver.wal.ProtobufLogWriter  
  315.         </value>  
  316.         <description>The WAL file writer implementation.</description>  
  317.     </property>  
  318.     <!-- regionServer的全局memstore的大小，超过该大小会触发flush到磁盘的操作,默认是堆大小的40%,而且regionserver级别的   
  319.         flush会阻塞客户端读写 -->  
  320.     <property>  
  321.         <name>hbase.regionserver.global.memstore.size</name>  
  322.         <value></value>  
  323.         <description>Maximum size of all memstores in a region server before  
  324.             new  
  325.             updates are blocked and flushes are forced. Defaults to 40% of heap (0.4).  
  326.             Updates are blocked and flushes are forced until size of all  
  327.             memstores  
  328.             in a region server hits  
  329.             hbase.regionserver.global.memstore.size.lower.limit.  
  330.             The default value in this configuration has been intentionally left  
  331.             emtpy in order to  
  332.             honor the old hbase.regionserver.global.memstore.upperLimit property if  
  333.             present.  
  334.         </description>  
  335.     </property>  
  336.     <!--可以理解为一个安全的设置，有时候集群的“写负载”非常高，写入量一直超过flush的量，这时，我们就希望memstore不要超过一定的安全设置。   
  337.         在这种情况下，写操作就要被阻塞一直到memstore恢复到一个“可管理”的大小, 这个大小就是默认值是堆大小 * 0.4 * 0.95，也就是当regionserver级别   
  338.         的flush操作发送后,会阻塞客户端写,一直阻塞到整个regionserver级别的memstore的大小为 堆大小 * 0.4 *0.95为止 -->  
  339.     <property>  
  340.         <name>hbase.regionserver.global.memstore.size.lower.limit</name>  
  341.         <value></value>  
  342.         <description>Maximum size of all memstores in a region server before  
  343.             flushes are forced.  
  344.             Defaults to 95% of hbase.regionserver.global.memstore.size (0.95).  
  345.             A 100% value for this value causes the minimum possible flushing to  
  346.             occur when updates are  
  347.             blocked due to memstore limiting.  
  348.             The default value in this configuration has been intentionally left  
  349.             emtpy in order to  
  350.             honor the old hbase.regionserver.global.memstore.lowerLimit property if  
  351.             present.  
  352.         </description>  
  353.     </property>  
  354.     <!-- 内存中的文件在自动刷新之前能够存活的最长时间，默认是1h -->  
  355.     <property>  
  356.         <name>hbase.regionserver.optionalcacheflushinterval</name>  
  357.         <value>3600000</value>  
  358.         <description>  
  359.             Maximum amount of time an edit lives in memory before being automatically  
  360.             flushed.  
  361.             Default 1 hour. Set it to 0 to disable automatic flushing.  
  362.         </description>  
  363.     </property>  
  364.     <property>  
  365.         <name>hbase.regionserver.catalog.timeout</name>  
  366.         <value>600000</value>  
  367.         <description>Timeout value for the Catalog Janitor from the  
  368.             regionserver to META.</description>  
  369.     </property>  
  370.     <!-- 当使用dns的时候，regionServer用来上报IP地址的网络接口名字 -->  
  371.     <property>  
  372.         <name>hbase.regionserver.dns.interface</name>  
  373.         <value>default</value>  
  374.         <description>The name of the Network Interface from which a region  
  375.             server  
  376.             should report its IP address.  
  377.         </description>  
  378.     </property>  
  379.     <!-- 当使用DNS的时候，RegionServer使用的DNS的域名或者IP 地址，RegionServer用它来确定和master用来进行通讯的域名 -->  
  380.     <property>  
  381.         <name>hbase.regionserver.dns.nameserver</name>  
  382.         <value>default</value>  
  383.         <description>The host name or IP address of the name server (DNS)  
  384.             which a region server should use to determine the host name used by  
  385.             the  
  386.             master for communication and display purposes.  
  387.         </description>  
  388.     </property>  
  389.     <!-- region在切分的时候的默认切分策略 -->  
  390.     <property>  
  391.         <name>hbase.regionserver.region.split.policy</name>  
  392.         <value>org.apache.hadoop.hbase.regionserver.IncreasingToUpperBoundRegionSplitPolicy  
  393.         </value>  
  394.         <description>  
  395.             A split policy determines when a region should be split. The various  
  396.             other split policies that  
  397.             are available currently are ConstantSizeRegionSplitPolicy,  
  398.             DisabledRegionSplitPolicy,  
  399.             DelimitedKeyPrefixRegionSplitPolicy, KeyPrefixRegionSplitPolicy etc.  
  400.         </description>  
  401.     </property>  
  402.     <!-- 当某个HRegionServer上的region到达这个限制时，不会在进行region切分，也就是一个HRegionServer默认最大允许有1000个region -->  
  403.     <property>  
  404.         <name>hbase.regionserver.regionSplitLimit</name>  
  405.         <value>1000</value>  
  406.         <description>  
  407.             Limit for the number of regions after which no more region splitting  
  408.             should take place.  
  409.             This is not hard limit for the number of regions but acts as a guideline  
  410.             for the regionserver  
  411.             to stop splitting after a certain limit. Default is set to 1000.  
  412.         </description>  
  413.     </property>  
  415.     <!-- zk sesscion超时时间 -->  
  416.     <property>  
  417.         <name>zookeeper.session.timeout</name>  
  418.         <value>90000</value>  
  419.         <description>ZooKeeper session timeout in milliseconds. It is used in  
  420.             two different ways.  
  421.             First, this value is used in the ZK client that HBase uses to connect to  
  422.             the ensemble.  
  423.             It is also used by HBase when it starts a ZK server and it is passed as  
  424.             the 'maxSessionTimeout'. See  
  425.             http://hadoop.apache.org/zookeeper/docs/current/zookeeperProgrammers.html#ch_zkSessions.  
  426.             For example, if a HBase region server connects to a ZK ensemble  
  427.             that's also managed by HBase, then the  
  428.             session timeout will be the one specified by this configuration. But, a  
  429.             region server that connects  
  430.             to an ensemble managed with a different configuration will be subjected  
  431.             that ensemble's maxSessionTimeout. So,  
  432.             even though HBase might propose using 90 seconds, the ensemble can have a  
  433.             max timeout lower than this and  
  434.             it will take precedence. The current default that ZK ships with is 40  
  435.             seconds, which is lower than HBase's.  
  436.         </description>  
  437.     </property>  
  438.     <!-- hbase在zk上默认的根目录 -->  
  439.     <property>  
  440.         <name>zookeeper.znode.parent</name>  
  441.         <value>/hbase</value>  
  442.         <description>Root ZNode for HBase in ZooKeeper. All of HBase's  
  443.             ZooKeeper  
  444.             files that are configured with a relative path will go under this node.  
  445.             By default, all of HBase's ZooKeeper file path are configured with a  
  446.             relative path, so they will all go under this directory unless  
  447.             changed.  
  448.         </description>  
  449.     </property>  
  450.     <!-- hbase在zk上的节点路径 -->  
  451.     <property>  
  452.         <name>zookeeper.znode.rootserver</name>  
  453.         <value>root-region-server</value>  
  454.         <description>Path to ZNode holding root region location. This is  
  455.             written by  
  456.             the master and read by clients and region servers. If a relative path is  
  457.             given, the parent folder will be ${zookeeper.znode.parent}. By  
  458.             default,  
  459.             this means the root location is stored at /hbase/root-region-server.  
  460.         </description>  
  461.     </property>  
  462.     <!-- hbase在zk上节点使用的权限 -->  
  463.     <property>  
  464.         <name>zookeeper.znode.acl.parent</name>  
  465.         <value>acl</value>  
  466.         <description>Root ZNode for access control lists.</description>  
  467.     </property>  
  468.     <property>  
  469.         <name>hbase.zookeeper.dns.interface</name>  
  470.         <value>default</value>  
  471.         <description>The name of the Network Interface from which a ZooKeeper  
  472.             server  
  473.             should report its IP address.  
  474.         </description>  
  475.     </property>  
  476.     <property>  
  477.         <name>hbase.zookeeper.dns.nameserver</name>  
  478.         <value>default</value>  
  479.         <description>The host name or IP address of the name server (DNS)  
  480.             which a ZooKeeper server should use to determine the host name used  
  481.             by the  
  482.             master for communication and display purposes.  
  483.         </description>  
  484.     </property>  
  485.     <!-- zk的使用端口 -->  
  486.     <property>  
  487.         <name>hbase.zookeeper.peerport</name>  
  488.         <value>2888</value>  
  489.         <description>Port used by ZooKeeper peers to talk to each other.  
  490.             See  
  491.             http://hadoop.apache.org/zookeeper/docs/r3.1.1/zookeeperStarted.html#sc_RunningReplicatedZooKeeper  
  492.             for more information.  
  493.         </description>  
  494.     </property>  
  495.     <!-- zk直接执行leader选举时通讯的端口 -->  
  496.     <property>  
  497.         <name>hbase.zookeeper.leaderport</name>  
  498.         <value>3888</value>  
  499.         <description>Port used by ZooKeeper for leader election.  
  500.             See  
  501.             http://hadoop.apache.org/zookeeper/docs/r3.1.1/zookeeperStarted.html#sc_RunningReplicatedZooKeeper  
  502.             for more information.  
  503.         </description>  
  504.     </property>  
  505.     <!-- zk是否支持多重更新 -->  
  506.     <property>  
  507.         <name>hbase.zookeeper.useMulti</name>  
  508.         <value>true</value>  
  509.         <description>Instructs HBase to make use of ZooKeeper's multi-update  
  510.             functionality.  
  511.             This allows certain ZooKeeper operations to complete more quickly and  
  512.             prevents some issues  
  513.             with rare Replication failure scenarios (see the release note of  
  514.             HBASE-2611 for an example).  
  515.             IMPORTANT: only set this to true if all ZooKeeper servers in the cluster are on  
  516.             version 3.4+  
  517.             and will not be downgraded. ZooKeeper versions before 3.4 do not support  
  518.             multi-update and  
  519.             will not fail gracefully if multi-update is invoked (see ZOOKEEPER-1495).  
  520.         </description>  
  521.     </property>  
  522.     <!-- 是否允许HBaseConfiguration去读取zk的配置文件，不清楚意义是什么? -->  
  523.     <property>  
  524.         <name>hbase.config.read.zookeeper.config</name>  
  525.         <value>false</value>  
  526.         <description>  
  527.             Set to true to allow HBaseConfiguration to read the  
  528.             zoo.cfg file for ZooKeeper properties. Switching this to true  
  529.             is not recommended, since the functionality of reading ZK  
  530.             properties from a zoo.cfg file has been deprecated.  
  531.         </description>  
  532.     </property>  
  533.     <property>  
  534.         <name>hbase.zookeeper.property.initLimit</name>  
  535.         <value>10</value>  
  536.         <description>Property from ZooKeeper's config zoo.cfg.  
  537.             The number of ticks that the initial synchronization phase can take.  
  538.         </description>  
  539.     </property>  
  540.     <property>  
  541.         <name>hbase.zookeeper.property.syncLimit</name>  
  542.         <value>5</value>  
  543.         <description>Property from ZooKeeper's config zoo.cfg.  
  544.             The number of ticks that can pass between sending a request and getting  
  545.             an  
  546.             acknowledgment.  
  547.         </description>  
  548.     </property>  
  549.     <property>  
  550.         <name>hbase.zookeeper.property.dataDir</name>  
  551.         <value>${hbase.tmp.dir}/zookeeper</value>  
  552.         <description>Property from ZooKeeper's config zoo.cfg.  
  553.             The directory where the snapshot is stored.  
  554.         </description>  
  555.     </property>  
  556.     <property>  
  557.         <name>hbase.zookeeper.property.clientPort</name>  
  558.         <value>2181</value>  
  559.         <description>Property from ZooKeeper's config zoo.cfg.  
  560.             The port at which the clients will connect.  
  561.         </description>  
  562.     </property>  
  563.     <property>  
  564.         <name>hbase.zookeeper.property.maxClientCnxns</name>  
  565.         <value>300</value>  
  566.         <description>Property from ZooKeeper's config zoo.cfg.  
  567.             Limit on number of concurrent connections (at the socket level) that a  
  568.             single client, identified by IP address, may make to a single member  
  569.             of  
  570.             the ZooKeeper ensemble. Set high to avoid zk connection issues running  
  571.             standalone and pseudo-distributed.  
  572.         </description>  
  573.     </property>  
  575.     <!--Client configurations -->  
  576.     <!-- hbase客户端每次 写缓冲的大小(也就是客户端批量提交到server端)，这块大小会同时占用客户端和服务端，缓冲区更大可以减少RPC次数，但是更大意味着内存占用更多 -->  
  577.     <property>  
  578.         <name>hbase.client.write.buffer</name>  
  579.         <value>2097152</value>  
  580.         <description>Default size of the HTable client write buffer in bytes.  
  581.             A bigger buffer takes more memory -- on both the client and server  
  582.             side since server instantiates the passed write buffer to process  
  583.             it -- but a larger buffer size reduces the number of RPCs made.  
  584.             For an estimate of server-side memory-used, evaluate  
  585.             hbase.client.write.buffer * hbase.regionserver.handler.count  
  586.         </description>  
  587.     </property>  
  588.     <!-- 在hbase发生请求失败的情况下,每次重试的等待时间 ,如果某段时间网络持续不好,重试会一直发生,如果还是连不上,就会放弃连接,在重试的过程中,会阻塞其它线程来抢锁,如果长时间的超时会导致业务处理的阻塞 -->  
  589.     <property>  
  590.         <name>hbase.client.pause</name>  
  591.         <value>100</value>  
  592.         <description>General client pause value. Used mostly as value to wait  
  593.             before running a retry of a failed get, region lookup, etc.  
  594.             See hbase.client.retries.number for description of how we backoff from  
  595.             this initial pause amount and how this pause works w/ retries.  
  596.         </description>  
  597.     </property>  
  598.     <!--重试次数，如果连不上或者fail，会重试 -->  
  599.     <property>  
  600.         <name>hbase.client.retries.number</name>  
  601.         <value>35</value>  
  602.         <description>Maximum retries. Used as maximum for all retryable  
  603.             operations such as the getting of a cell's value, starting a row  
  604.             update,  
  605.             etc. Retry interval is a rough function based on hbase.client.pause. At  
  606.             first we retry at this interval but then with backoff, we pretty  
  607.             quickly reach  
  608.             retrying every ten seconds. See HConstants#RETRY_BACKOFF for how the backup  
  609.             ramps up. Change this setting and hbase.client.pause to suit your  
  610.             workload.  
  611.         </description>  
  612.     </property>  
  613.     <!-- 单个Htable实例发送给集群的最大任务数,也就是同一个实例最大的并发数 -->  
  614.     <property>  
  615.         <name>hbase.client.max.total.tasks</name>  
  616.         <value>100</value>  
  617.         <description>The maximum number of concurrent tasks a single HTable  
  618.             instance will  
  619.             send to the cluster.  
  620.         </description>  
  621.     </property>  
  622.     <!-- 单个Htable实例发给regionServer的最大的任务并发数 -->  
  623.     <property>  
  624.         <name>hbase.client.max.perserver.tasks</name>  
  625.         <value>5</value>  
  626.         <description>The maximum number of concurrent tasks a single HTable  
  627.             instance will  
  628.             send to a single region server.  
  629.         </description>  
  630.     </property>  
  631.     <!-- 客户端到一个region的最大连接数，也就是说如果一个客户端有超过配置项值到某个region的连接，后面的请求会被阻塞 -->  
  632.     <property>  
  633.         <name>hbase.client.max.perregion.tasks</name>  
  634.         <value>1</value>  
  635.         <description>The maximum number of concurrent connections the client  
  636.             will  
  637.             maintain to a single Region. That is, if there is already  
  638.             hbase.client.max.perregion.tasks writes in progress for this region,  
  639.             new puts  
  640.             won't be sent to this region until some writes finishes.  
  641.         </description>  
  642.     </property>  
  643.     <!-- 在执行hbase scan操作的时候，客户端缓存的行数，设置小意味着更多的rpc次数，设置大比较吃内存 -->  
  644.     <property>  
  645.         <name>hbase.client.scanner.caching</name>  
  646.         <value>2147483647</value>  
  647.         <description>Number of rows that we try to fetch when calling next  
  648.             on a scanner if it is not served from (local, client) memory. This  
  649.             configuration  
  650.             works together with hbase.client.scanner.max.result.size to try and use  
  651.             the  
  652.             network efficiently. The default value is Integer.MAX_VALUE by default so  
  653.             that  
  654.             the network will fill the chunk size defined by  
  655.             hbase.client.scanner.max.result.size  
  656.             rather than be limited by a particular number of rows since the size of  
  657.             rows varies  
  658.             table to table. If you know ahead of time that you will not require more  
  659.             than a certain  
  660.             number of rows from a scan, this configuration should be set to that row  
  661.             limit via  
  662.             Scan#setCaching. Higher caching values will enable faster scanners but will eat up  
  663.             more  
  664.             memory and some calls of next may take longer and longer times when the  
  665.             cache is empty.  
  666.             Do not set this value such that the time between invocations is greater  
  667.             than the scanner  
  668.             timeout; i.e. hbase.client.scanner.timeout.period  
  669.         </description>  
  670.     </property>  
  671.     <!--一个KeyValue实例的最大大小，这是存储文件中一个entry的容量上限,因为一个KeyValue是不能分割的， 所有可以避免因为数据过大导致region不可分割 -->  
  672.     <property>  
  673.         <name>hbase.client.keyvalue.maxsize</name>  
  674.         <value>10485760</value>  
  675.         <description>Specifies the combined maximum allowed size of a KeyValue  
  676.             instance. This is to set an upper boundary for a single entry saved  
  677.             in a  
  678.             storage file. Since they cannot be split it helps avoiding that a region  
  679.             cannot be split any further because the data is too large. It seems  
  680.             wise  
  681.             to set this to a fraction of the maximum region size. Setting it to  
  682.             zero  
  683.             or less disables the check.  
  684.         </description>  
  685.     </property>  
  686.     <!-- scan操作中单次rpc的超时时间(比较重要的参数) -->  
  687.     <property>  
  688.         <name>hbase.client.scanner.timeout.period</name>  
  689.         <value>60000</value>  
  690.         <description>Client scanner lease period in milliseconds.  
  691.         </description>  
  692.     </property>  
  693.     <property>  
  694.         <name>hbase.client.localityCheck.threadPoolSize</name>  
  695.         <value>2</value>  
  696.     </property>  
  697.     <!--Miscellaneous configuration -->  
  698.     <property>  
  699.         <name>hbase.bulkload.retries.number</name>  
  700.         <value>10</value>  
  701.         <description>Maximum retries. This is maximum number of iterations  
  702.             to atomic bulk loads are attempted in the face of splitting operations  
  703.             0 means never give up.  
  704.         </description>  
  705.     </property>  
  706.     <property>  
  707.         <name>hbase.balancer.period</name>  
  708.         <value>300000</value>  
  709.         <description>Period at which the region balancer runs in the Master.  
  710.         </description>  
  711.     </property>  
  712.     <property>  
  713.         <name>hbase.normalizer.period</name>  
  714.         <value>1800000</value>  
  715.         <description>Period at which the region normalizer runs in the Master.  
  716.         </description>  
  717.     </property>  
  718.     <!-- HRegion负载迁移的时候的一个配置参数,具体怎么用可看HMaster里面的负载迁移的源代码 -->  
  719.     <property>  
  720.         <name>hbase.regions.slop</name>  
  721.         <value>0.2</value>  
  722.         <description>Rebalance if any regionserver has average + (average *  
  723.             slop) regions.</description>  
  724.     </property>  
  725.     <!-- 每次线程唤醒的周期 -->  
  726.     <property>  
  727.         <name>hbase.server.thread.wakefrequency</name>  
  728.         <value>10000</value>  
  729.         <description>Time to sleep in between searches for work (in  
  730.             milliseconds).  
  731.             Used as sleep interval by service threads such as log roller.  
  732.         </description>  
  733.     </property>  
  734.     <property>  
  735.         <name>hbase.server.versionfile.writeattempts</name>  
  736.         <value>3</value>  
  737.         <description>  
  738.             How many time to retry attempting to write a version file  
  739.             before just aborting. Each attempt is seperated by the  
  740.             hbase.server.thread.wakefrequency milliseconds.  
  741.         </description>  
  742.     </property>  
  743.     <!-- 单个region里memstore的缓存大小，超过那么整个HRegion就会flush,默认128M -->  
  744.     <property>  
  745.         <name>hbase.hregion.memstore.flush.size</name>  
  746.         <value>134217728</value>  
  747.         <description>  
  748.             Memstore will be flushed to disk if size of the memstore  
  749.             exceeds this number of bytes. Value is checked by a thread that runs  
  750.             every hbase.server.thread.wakefrequency.  
  751.         </description>  
  752.     </property>  
  753.     <property>  
  754.         <name>hbase.hregion.percolumnfamilyflush.size.lower.bound</name>  
  755.         <value>16777216</value>  
  756.         <description>  
  757.             If FlushLargeStoresPolicy is used, then every time that we hit the  
  758.             total memstore limit, we find out all the column families whose  
  759.             memstores  
  760.             exceed this value, and only flush them, while retaining the others whose  
  761.             memstores are lower than this limit. If none of the families have  
  762.             their  
  763.             memstore size more than this, all the memstores will be flushed  
  764.             (just as usual). This value should be less than half of the total memstore  
  765.             threshold (hbase.hregion.memstore.flush.size).  
  766.         </description>  
  767.     </property>  
  768.     <!--当一个 region 中的 memstore 的大小大于这个值的时候，我们又触发 了 close.会先运行“pre-flush”操作，清理这个需要关闭的   
  769.         memstore，然后 将这个 region 下线。当一个 region 下线了，我们无法再进行任何写操作。 如果一个 memstore 很大的时候，flush   
  770.         操作会消耗很多时间。"pre-flush" 操作意味着在 region 下线之前，会先把 memstore 清空。这样在最终执行 close 操作的时候，flush   
  771.         操作会很快。 -->  
  772.     <property>  
  773.         <name>hbase.hregion.preclose.flush.size</name>  
  774.         <value>5242880</value>  
  775.         <description>  
  776.             If the memstores in a region are this size or larger when we go  
  777.             to close, run a "pre-flush" to clear out memstores before we put up  
  778.             the region closed flag and take the region offline. On close,  
  779.             a flush is run under the close flag to empty memory. During  
  780.             this time the region is offline and we are not taking on any writes.  
  781.             If the memstore content is large, this flush could take a long time to  
  782.             complete. The preflush is meant to clean out the bulk of the memstore  
  783.             before putting up the close flag and taking the region offline so the  
  784.             flush that runs under the close flag has little to do.  
  785.         </description>  
  786.     </property>  
  787.     <!-- 当一个HRegion上的memstore的大小满足hbase.hregion.memstore.block.multiplier *   
  788.         hbase.hregion.memstore.flush.size, 这个HRegion会执行flush操作并阻塞对该HRegion的写入 -->  
  789.     <property>  
  790.         <name>hbase.hregion.memstore.block.multiplier</name> 
  791.         <value>4</value>  
  792.         <description>  
  793.             Block updates if memstore has hbase.hregion.memstore.block.multiplier  
  794.             times hbase.hregion.memstore.flush.size bytes. Useful preventing  
  795.             runaway memstore during spikes in update traffic. Without an  
  796.             upper-bound, memstore fills such that when it flushes the  
  797.             resultant flush files take a long time to compact or split, or  
  798.             worse, we OOME.  
  799.         </description>  
  800.     </property>  
  801.     <!-- 设置为true,有效减少在高并发写时候的内存碎片 -->  
  802.     <property>  
  803.         <name>hbase.hregion.memstore.mslab.enabled</name>  
  804.         <value>true</value>  
  805.         <description>  
  806.             Enables the MemStore-Local Allocation Buffer,  
  807.             a feature which works to prevent heap fragmentation under  
  808.             heavy write loads. This can reduce the frequency of stop-the-world  
  809.             GC pauses on large heaps.  
  810.         </description>  
  811.     </property>  
  812.     <!--HStoreFile最大的大小，当某个region的某个列族超过这个大小会进行region拆分 -->  
  813.     <property>  
  814.         <name>hbase.hregion.max.filesize</name>  
  815.         <value>10737418240</value>  
  816.         <description>  
  817.             Maximum HStoreFile size. If any one of a column families' HStoreFiles has  
  818.             grown to exceed this value, the hosting HRegion is split in two.  
  819.         </description>  
  820.     </property>  
  821.     <!-- 一个region进行 major compaction合并的周期,在这个点的时候， 这个region下的所有hfile会进行合并,默认是7天,major   
  822.         compaction非常耗资源,建议生产关闭(设置为0)，在应用空闲时间手动触发 -->  
  823.     <property>  
  824.         <name>hbase.hregion.majorcompaction</name>  
  825.         <value>604800000</value>  
  826.         <description>The time (in miliseconds) between 'major' compactions of  
  827.             all  
  828.             HStoreFiles in a region. Default: Set to 7 days. Major compactions tend to  
  829.             happen exactly when you need them least so enable them such that they  
  830.             run at  
  831.             off-peak for your deploy; or, since this setting is on a periodicity that is  
  832.             unlikely to match your loading, run the compactions via an external  
  833.             invocation out of a cron job or some such.  
  834.         </description>  
  835.     </property>  
  836.     <!-- 一个抖动比例，意思是说上一个参数设置是7天进行一次合并，也可以有50%的抖动比例 -->  
  837.     <property>  
  838.         <name>hbase.hregion.majorcompaction.jitter</name>  
  839.         <value>0.50</value>  
  840.         <description>Jitter outer bound for major compactions.  
  841.             On each regionserver, we multiply the hbase.region.majorcompaction  
  842.             interval by some random fraction that is inside the bounds of this  
  843.             maximum. We then add this + or - product to when the next  
  844.             major compaction is to run. The idea is that major compaction  
  845.             does happen on every regionserver at exactly the same time. The  
  846.             smaller this number, the closer the compactions come together.  
  847.         </description>  
  848.     </property>  
  849.     <!-- 一个store里面允许存的hfile的个数，超过这个个数会被写到新的一个hfile里面 也即是每个region的每个列族对应的memstore在fulsh为hfile的时候，默认情况下当达到3个hfile的时候就会   
  850.         对这些文件进行合并重写为一个新文件，设置个数越大可以减少触发合并的时间，但是每次合并的时间就会越长 -->  
  851.     <property>  
  852.         <name>hbase.hstore.compactionThreshold</name>  
  853.         <value>3</value>  
  854.         <description>  
  855.             If more than this number of HStoreFiles in any one HStore  
  856.             (one HStoreFile is written per flush of memstore) then a compaction  
  857.             is run to rewrite all HStoreFiles files as one. Larger numbers  
  858.             put off compaction but when it runs, it takes longer to complete.  
  859.         </description>  
  860.     </property>  
  861.     <!-- 执行flush操作的线程数,设置小了刷新操作会排队，大了会增加底层hdfs的负载压力 -->  
  862.     <property>  
  863.         <name>hbase.hstore.flusher.count</name>  
  864.         <value>2</value>  
  865.         <description>  
  866.             The number of flush threads. With less threads, the memstore flushes  
  867.             will be queued. With  
  868.             more threads, the flush will be executed in parallel, increasing the hdfs  
  869.             load. This can  
  870.             lead as well to more compactions.  
  871.         </description>  
  872.     </property>  
  873.     <!-- 每个store阻塞更新请求的阀值,表示如果当前hstore中文件数大于该值，系统将会强制执行compaction操作进行文件合并， 合并的过程会阻塞整个hstore的写入,这样有个好处是避免compaction操作赶不上Hfile文件的生成速率 -->  
  874.     <property>  
  875.         <name>hbase.hstore.blockingStoreFiles</name>  
  876.         <value>10</value>  
  877.         <description>  
  878.             If more than this number of StoreFiles in any one Store  
  879.             (one StoreFile is written per flush of MemStore) then updates are  
  880.             blocked for this HRegion until a compaction is completed, or  
  881.             until hbase.hstore.blockingWaitTime has been exceeded.  
  882.         </description>  
  883.     </property>  
  884.     <!-- 每个store阻塞更新请求的超时时间，如果超过这个时间合并操作还未完成，阻塞也会取消 -->  
  885.     <property>  
  886.         <name>hbase.hstore.blockingWaitTime</name>  
  887.         <value>90000</value>  
  888.         <description>  
  889.             The time an HRegion will block updates for after hitting the StoreFile  
  890.             limit defined by hbase.hstore.blockingStoreFiles.  
  891.             After this time has elapsed, the HRegion will stop blocking updates even  
  892.             if a compaction has not been completed.  
  893.         </description>  
  894.     </property>  
  895.     <!-- 每个minor compaction操作的 允许的最大hfile文件上限 -->  
  896.     <property>  
  897.         <name>hbase.hstore.compaction.max</name>  
  898.         <value>10</value>  
  899.         <description>Max number of HStoreFiles to compact per 'minor'  
  900.             compaction.</description>  
  901.     </property>  
  902.     <!-- 在执行compaction操作的过程中，每次读取hfile文件的keyValue个数 -->  
  903.     <property>  
  904.         <name>hbase.hstore.compaction.kv.max</name>  
  905.         <value>10</value>  
  906.         <description>How many KeyValues to read and then write in a batch when  
  907.             flushing  
  908.             or compacting. Do less if big KeyValues and problems with OOME.  
  909.             Do more if wide, small rows.  
  910.         </description>  
  911.     </property>  
  912.     <property>  
  913.         <name>hbase.hstore.time.to.purge.deletes</name>  
  914.         <value>0</value>  
  915.         <description>The amount of time to delay purging of delete markers  
  916.             with future timestamps. If  
  917.             unset, or set to 0, all delete markers, including those with future  
  918.             timestamps, are purged  
  919.             during the next major compaction. Otherwise, a delete marker is kept until  
  920.             the major compaction  
  921.             which occurs after the marker's timestamp plus the value of this setting,  
  922.             in milliseconds.  
  923.         </description>  
  924.     </property>  
  925.     <property>  
  926.         <name>hbase.storescanner.parallel.seek.enable</name>  
  927.         <value>false</value>  
  928.         <description>  
  929.             Enables StoreFileScanner parallel-seeking in StoreScanner,  
  930.             a feature which can reduce response latency under special conditions.  
  931.         </description>  
  932.     </property>  
  933.     <property>  
  934.         <name>hbase.storescanner.parallel.seek.threads</name>  
  935.         <value>10</value>  
  936.         <description>  
  937.             The default thread pool size if parallel-seeking feature enabled.  
  938.         </description>  
  939.     </property>  
  940.     <!--LRUBlockCache块缓存的大小，默认为堆大小的40% -->  
  941.     <property>  
  942.         <name>hfile.block.cache.size</name>  
  943.         <value>0.4</value>  
  944.         <description>Percentage of maximum heap (-Xmx setting) to allocate to  
  945.             block cache  
  946.             used by HFile/StoreFile. Default of 0.4 means allocate 40%.  
  947.             Set to 0 to disable but it's not recommended; you need at least  
  948.             enough cache to hold the storefile indices.  
  949.         </description>  
  950.     </property>  
  951.     <property>  
  952.         <name>hfile.block.index.cacheonwrite</name>  
  953.         <value>false</value>  
  954.         <description>This allows to put non-root multi-level index blocks into  
  955.             the block  
  956.             cache at the time the index is being written.  
  957.         </description>  
  958.     </property>  
  959.     <property>  
  960.         <name>hfile.index.block.max.size</name>  
  961.         <value>131072</value>  
  962.         <description>When the size of a leaf-level, intermediate-level, or  
  963.             root-level  
  964.             index block in a multi-level block index grows to this size, the  
  965.             block is written out and a new block is started.  
  966.         </description>  
  967.     </property>  
  968.     <!--bucketcache的工作模式,默认有3种可选择，heap,offheap,file。其中heap由jvm分配内存存储，offheap   
  969.         由操作系统分配内存存储 -->  
  970.     <property>  
  971.         <name>hbase.bucketcache.ioengine</name>  
  972.         <value></value>  
  973.         <description>Where to store the contents of the bucketcache. One of:  
  974.             heap,  
  975.             offheap, or file. If a file, set it to file:PATH_TO_FILE. See  
  976.             http://hbase.apache.org/book.html#offheap.blockcache for more  
  977.             information.  
  978.         </description>  
  979.     </property>  
  980.     <!-- 默认为true,意思是combinedcache里面包括了LRU和bucketcache -->  
  981.     <property>  
  982.         <name>hbase.bucketcache.combinedcache.enabled</name>  
  983.         <value>true</value>  
  984.         <description>Whether or not the bucketcache is used in league with the  
  985.             LRU  
  986.             on-heap block cache. In this mode, indices and blooms are kept in the LRU  
  987.             blockcache and the data blocks are kept in the bucketcache.  
  988.         </description>  
  989.     </property>  
  990.     <!-- 就是bucketcache大小,如果配置的值在0-1直接,表示占用堆内存的百分比,或者配置XXMB也可 -->  
  991.     <property>  
  992.         <name>hbase.bucketcache.size</name>  
  993.         <value></value>  
  994.         <description>A float that EITHER represents a percentage of total heap  
  995.             memory size to give to the cache (if &lt; 1.0) OR, it is the total capacity in  megabytes of BucketCache. Default: 0.0  
  996.         </description>  
  997.     </property>  
  998.     <property>  
  999.         <name>hbase.bucketcache.sizes</name>  
  1000.         <value></value>  
  1001.         <description>A comma-separated list of sizes for buckets for the  
  1002.             bucketcache.  
  1003.             Can be multiple sizes. List block sizes in order from smallest to  
  1004.             largest.  
  1005.             The sizes you use will depend on your data access patterns.  
  1006.             Must be a multiple of 1024 else you will run into  
  1007.             'java.io.IOException: Invalid HFile block magic' when you go to read from cache.  
  1008.             If you specify no values here, then you pick up the default bucketsizes  
  1009.             set  
  1010.             in code (See BucketAllocator#DEFAULT_BUCKET_SIZES).  
  1011.         </description>  
  1012.     </property>  
  1013.     <property>  
  1014.         <name>hfile.format.version</name>  
  1015.         <value>3</value>  
  1016.         <description>The HFile format version to use for new files.  
  1017.             Version 3 adds support for tags in hfiles (See  
  1018.             http://hbase.apache.org/book.html#hbase.tags).  
  1019.             Distributed Log Replay requires that tags are enabled. Also see the  
  1020.             configuration  
  1021.             'hbase.replication.rpc.codec'.  
  1022.         </description>  
  1023.     </property>  
  1024.     <property>  
  1025.         <name>hfile.block.bloom.cacheonwrite</name>  
  1026.         <value>false</value>  
  1027.         <description>Enables cache-on-write for inline blocks of a compound  
  1028.             Bloom filter.</description>  
  1029.     </property>  
  1030.     <property>  
  1031.         <name>io.storefile.bloom.block.size</name>  
  1032.         <value>131072</value>  
  1033.         <description>The size in bytes of a single block ("chunk") of a  
  1034.             compound Bloom  
  1035.             filter. This size is approximate, because Bloom blocks can only be  
  1036.             inserted at data block boundaries, and the number of keys per data  
  1037.             block varies.  
  1038.         </description>  
  1039.     </property>  
  1040.     <property>  
  1041.         <name>hbase.rs.cacheblocksonwrite</name>  
  1042.         <value>false</value>  
  1043.         <description>Whether an HFile block should be added to the block cache  
  1044.             when the  
  1045.             block is finished.  
  1046.         </description>  
  1047.     </property>  
  1048.     <!-- 单次rpc请求的超时时间,如果某次RPC时间超过该值，客户端就会主动关闭socket -->  
  1049.     <property>  
  1050.         <name>hbase.rpc.timeout</name>  
  1051.         <value>60000</value>  
  1052.         <description>This is for the RPC layer to define how long  
  1053.             (millisecond) HBase client applications  
  1054.             take for a remote call to time out. It uses pings to check connections  
  1055.             but will eventually throw a TimeoutException.  
  1056.         </description>  
  1057.     </property>  
  1058.     <!-- 该参数表示HBase客户端发起一次数据操作(一次操作可能有多次rpc)直至得到响应之间总的超时时间 -->  
  1059.     <property>  
  1060.         <name>hbase.client.operation.timeout</name>  
  1061.         <value>1200000</value>  
  1062.         <description>Operation timeout is a top-level restriction  
  1063.             (millisecond) that makes sure a  
  1064.             blocking operation in Table will not be blocked more than this. In each  
  1065.             operation, if rpc  
  1066.             request fails because of timeout or other reason, it will retry until  
  1067.             success or throw  
  1068.             RetriesExhaustedException. But if the total time being blocking reach the operation timeout  
  1069.             before retries exhausted, it will break early and throw  
  1070.             SocketTimeoutException.  
  1071.         </description>  
  1072.     </property>  
  1073.     <property>  
  1074.         <name>hbase.cells.scanned.per.heartbeat.check</name>  
  1075.         <value>10000</value>  
  1076.         <description>The number of cells scanned in between heartbeat checks.  
  1077.             Heartbeat  
  1078.             checks occur during the processing of scans to determine whether or not the  
  1079.             server should stop scanning in order to send back a heartbeat message  
  1080.             to the  
  1081.             client. Heartbeat messages are used to keep the client-server connection  
  1082.             alive  
  1083.             during long running scans. Small values mean that the heartbeat checks will  
  1084.             occur more often and thus will provide a tighter bound on the  
  1085.             execution time of  
  1086.             the scan. Larger values mean that the heartbeat checks occur less  
  1087.             frequently  
  1088.         </description>  
  1089.     </property>  
  1090.     <property>  
  1091.         <name>hbase.rpc.shortoperation.timeout</name>  
  1092.         <value>10000</value>  
  1093.         <description>This is another version of "hbase.rpc.timeout". For those  
  1094.             RPC operation  
  1095.             within cluster, we rely on this configuration to set a short timeout  
  1096.             limitation  
  1097.             for short operation. For example, short rpc timeout for region server's  
  1098.             trying  
  1099.             to report to active master can benefit quicker master failover process.  
  1100.         </description>  
  1101.     </property>  
  1102.     <property>  
  1103.         <name>hbase.ipc.client.tcpnodelay</name>  
  1104.         <value>true</value>  
  1105.         <description>Set no delay on rpc socket connections. See  
  1106.             http://docs.oracle.com/javase/1.5.0/docs/api/java/net/Socket.html#getTcpNoDelay()  
  1107.         </description>  
  1108.     </property>  
  1109.     <property>  
  1110.         <name>hbase.regionserver.hostname</name>  
  1111.         <value></value>  
  1112.         <description>This config is for experts: don't set its value unless  
  1113.             you really know what you are doing.  
  1114.             When set to a non-empty value, this represents the (external facing)  
  1115.             hostname for the underlying server.  
  1116.             See https://issues.apache.org/jira/browse/HBASE-12954 for details.  
  1117.         </description>  
  1118.     </property>  
  1119.     <!-- The following properties configure authentication information for HBase   
  1120.         processes when using Kerberos security. There are no default values, included   
  1121.         here for documentation purposes -->  
  1122.     <property>  
  1123.         <name>hbase.master.keytab.file</name>  
  1124.         <value></value>  
  1125.         <description>Full path to the kerberos keytab file to use for logging  
  1126.             in  
  1127.             the configured HMaster server principal.  
  1128.         </description>  
  1129.     </property>  
  1130.     <property>  
  1131.         <name>hbase.master.kerberos.principal</name>  
  1132.         <value></value>  
  1133.         <description>Ex. "hbase/_HOST@EXAMPLE.COM". The kerberos principal  
  1134.             name  
  1135.             that should be used to run the HMaster process. The principal name should  
  1136.             be in the form: user/hostname@DOMAIN. If "_HOST" is used as the  
  1137.             hostname  
  1138.             portion, it will be replaced with the actual hostname of the running  
  1139.             instance.  
  1140.         </description>  
  1141.     </property>  
  1142.     <property>  
  1143.         <name>hbase.regionserver.keytab.file</name>  
  1144.         <value></value>  
  1145.         <description>Full path to the kerberos keytab file to use for logging  
  1146.             in  
  1147.             the configured HRegionServer server principal.  
  1148.         </description>  
  1149.     </property>  
  1150.     <property>  
  1151.         <name>hbase.regionserver.kerberos.principal</name>  
  1152.         <value></value>  
  1153.         <description>Ex. "hbase/_HOST@EXAMPLE.COM". The kerberos principal  
  1154.             name  
  1155.             that should be used to run the HRegionServer process. The principal name  
  1156.             should be in the form: user/hostname@DOMAIN. If "_HOST" is used as  
  1157.             the  
  1158.             hostname portion, it will be replaced with the actual hostname of the  
  1159.             running instance. An entry for this principal must exist in the file  
  1160.             specified in hbase.regionserver.keytab.file  
  1161.         </description>  
  1162.     </property>  
  1163.     <!-- Additional configuration specific to HBase security -->  
  1164.     <property>  
  1165.         <name>hadoop.policy.file</name>  
  1166.         <value>hbase-policy.xml</value>  
  1167.         <description>The policy configuration file used by RPC servers to make  
  1168.             authorization decisions on client requests. Only used when HBase  
  1169.             security is enabled.  
  1170.         </description>  
  1171.     </property>  
  1172.     <property>  
  1173.         <name>hbase.superuser</name>  
  1174.         <value></value>  
  1175.         <description>List of users or groups (comma-separated), who are  
  1176.             allowed  
  1177.             full privileges, regardless of stored ACLs, across the cluster.  
  1178.             Only used when HBase security is enabled.  
  1179.         </description>  
  1180.     </property>  
  1181.     <property>  
  1182.         <name>hbase.auth.key.update.interval</name>  
  1183.         <value>86400000</value>  
  1184.         <description>The update interval for master key for authentication  
  1185.             tokens  
  1186.             in servers in milliseconds. Only used when HBase security is enabled.  
  1187.         </description>  
  1188.     </property>  
  1189.     <property>  
  1190.         <name>hbase.auth.token.max.lifetime</name>  
  1191.         <value>604800000</value>  
  1192.         <description>The maximum lifetime in milliseconds after which an  
  1193.             authentication token expires. Only used when HBase security is  
  1194.             enabled.  
  1195.         </description>  
  1196.     </property>  
  1197.     <property>  
  1198.         <name>hbase.ipc.client.fallback-to-simple-auth-allowed</name>  
  1199.         <value>false</value>  
  1200.         <description>When a client is configured to attempt a secure  
  1201.             connection, but attempts to  
  1202.             connect to an insecure server, that server may instruct the client to  
  1203.             switch to SASL SIMPLE (unsecure) authentication. This setting controls  
  1204.             whether or not the client will accept this instruction from the  
  1205.             server.  
  1206.             When false (the default), the client will not allow the fallback to  
  1207.             SIMPLE  
  1208.             authentication, and will abort the connection.  
  1209.         </description>  
  1210.     </property>  
  1211.     <property>  
  1212.         <name>hbase.ipc.server.fallback-to-simple-auth-allowed</name>  
  1213.         <value>false</value>  
  1214.         <description>When a server is configured to require secure  
  1215.             connections, it will  
  1216.             reject connection attempts from clients using SASL SIMPLE (unsecure)  
  1217.             authentication.  
  1218.             This setting allows secure servers to accept SASL SIMPLE connections from  
  1219.             clients  
  1220.             when the client requests. When false (the default), the server will not  
  1221.             allow the fallback  
  1222.             to SIMPLE authentication, and will reject the connection. WARNING: This  
  1223.             setting should ONLY  
  1224.             be used as a temporary measure while converting clients over to secure  
  1225.             authentication. It  
  1226.             MUST BE DISABLED for secure operation.  
  1227.         </description>  
  1228.     </property>  
  1229.     <property>  
  1230.         <name>hbase.coprocessor.enabled</name>  
  1231.         <value>true</value>  
  1232.         <description>Enables or disables coprocessor loading. If 'false'  
  1233.             (disabled), any other coprocessor related configuration will be  
  1234.             ignored.  
  1235.         </description>  
  1236.     </property>  
  1237.     <property>  
  1238.         <name>hbase.coprocessor.user.enabled</name>  
  1239.         <value>true</value>  
  1240.         <description>Enables or disables user (aka. table) coprocessor  
  1241.             loading.  
  1242.             If 'false' (disabled), any table coprocessor attributes in table  
  1243.             descriptors will be ignored. If "hbase.coprocessor.enabled" is  
  1244.             'false'  
  1245.             this setting has no effect.  
  1246.         </description>  
  1247.     </property>  
  1248.     <property>  
  1249.         <name>hbase.coprocessor.region.classes</name>  
  1250.         <value></value>  
  1251.         <description>A comma-separated list of Coprocessors that are loaded by  
  1252.             default on all tables. For any override coprocessor method, these  
  1253.             classes  
  1254.             will be called in order. After implementing your own Coprocessor, just  
  1255.             put  
  1256.             it in HBase's classpath and add the fully qualified class name here.  
  1257.             A coprocessor can also be loaded on demand by setting  
  1258.             HTableDescriptor.  
  1259.         </description>  
  1260.     </property>  
  1261.     <property>  
  1262.         <name>hbase.rest.port</name>  
  1263.         <value>8080</value>  
  1264.         <description>The port for the HBase REST server.</description>  
  1265.     </property>  
  1266.     <property>  
  1267.         <name>hbase.rest.readonly</name>  
  1268.         <value>false</value>  
  1269.         <description>Defines the mode the REST server will be started in.  
  1270.             Possible values are:  
  1271.             false: All HTTP methods are permitted - GET/PUT/POST/DELETE.  
  1272.             true: Only the GET method is permitted.  
  1273.         </description>  
  1274.     </property>  
  1275.     <property>  
  1276.         <name>hbase.rest.threads.max</name>  
  1277.         <value>100</value>  
  1278.         <description>The maximum number of threads of the REST server thread  
  1279.             pool.  
  1280.             Threads in the pool are reused to process REST requests. This  
  1281.             controls the maximum number of requests processed concurrently.  
  1282.             It may help to control the memory used by the REST server to  
  1283.             avoid OOM issues. If the thread pool is full, incoming requests  
  1284.             will be queued up and wait for some free threads.  
  1285.         </description>  
  1286.     </property>  
  1287.     <property>  
  1288.         <name>hbase.rest.threads.min</name>  
  1289.         <value>2</value>  
  1290.         <description>The minimum number of threads of the REST server thread  
  1291.             pool.  
  1292.             The thread pool always has at least these number of threads so  
  1293.             the REST server is ready to serve incoming requests.  
  1294.         </description>  
  1295.     </property>  
  1296.     <property>  
  1297.         <name>hbase.rest.support.proxyuser</name>  
  1298.         <value>false</value>  
  1299.         <description>Enables running the REST server to support proxy-user  
  1300.             mode.</description>  
  1301.     </property>  
  1302.     <property skipInDoc="true">  
  1303.         <name>hbase.defaults.for.version</name>  
  1304.         <value>1.2.3</value>  
  1305.         <description>This defaults file was compiled for version  
  1306.             ${project.version}. This variable is used  
  1307.             to make sure that a user doesn't have an old version of  
  1308.             hbase-default.xml on the  
  1309.             classpath.  
  1310.         </description>  
  1311.     </property>  
  1312.     <property>  
  1313.         <name>hbase.defaults.for.version.skip</name>  
  1314.         <value>false</value>  
  1315.         <description>Set to true to skip the 'hbase.defaults.for.version'  
  1316.             check.  
  1317.             Setting this to true can be useful in contexts other than  
  1318.             the other side of a maven generation; i.e. running in an  
  1319.             ide. You'll want to set this boolean to true to avoid  
  1320.             seeing the RuntimException complaint: "hbase-default.xml file  
  1321.             seems to be for and old version of HBase (\${hbase.version}), this  
  1322.             version is X.X.X-SNAPSHOT"  
  1323.         </description>  
  1324.     </property>  
  1325.     <property>  
  1326.         <name>hbase.coprocessor.master.classes</name>  
  1327.         <value></value>  
  1328.         <description>A comma-separated list of  
  1329.             org.apache.hadoop.hbase.coprocessor.MasterObserver coprocessors that  
  1330.             are  
  1331.             loaded by default on the active HMaster process. For any implemented  
  1332.             coprocessor methods, the listed classes will be called in order.  
  1333.             After  
  1334.             implementing your own MasterObserver, just put it in HBase's classpath  
  1335.             and add the fully qualified class name here.  
  1336.         </description>  
  1337.     </property>  
  1338.     <property>  
  1339.         <name>hbase.coprocessor.abortonerror</name>  
  1340.         <value>true</value>  
  1341.         <description>Set to true to cause the hosting server (master or  
  1342.             regionserver)  
  1343.             to abort if a coprocessor fails to load, fails to initialize, or throws  
  1344.             an  
  1345.             unexpected Throwable object. Setting this to false will allow the server to  
  1346.             continue execution but the system wide state of the coprocessor in  
  1347.             question  
  1348.             will become inconsistent as it will be properly executing in only a  
  1349.             subset  
  1350.             of servers, so this is most useful for debugging only.  
  1351.         </description>  
  1352.     </property>  
  1353.     <property>  
  1354.         <name>hbase.online.schema.update.enable</name>  
  1355.         <value>true</value>  
  1356.         <description>Set true to enable online schema changes.</description>  
  1357.     </property>  
  1358.     <property>  
  1359.         <name>hbase.table.lock.enable</name>  
  1360.         <value>true</value>  
  1361.         <description>Set to true to enable locking the table in zookeeper for  
  1362.             schema change operations.  
  1363.             Table locking from master prevents concurrent schema modifications to  
  1364.             corrupt table  
  1365.             state.  
  1366.         </description>  
  1367.     </property>  
  1368.     <!-- hbase table单行row的最大大小 -->  
  1369.     <property>  
  1370.         <name>hbase.table.max.rowsize</name>  
  1371.         <value>1073741824</value>  
  1372.         <description>  
  1373.             Maximum size of single row in bytes (default is 1 Gb) for Get'ting  
  1374.             or Scan'ning without in-row scan flag set. If row size exceeds this  
  1375.             limit  
  1376.             RowTooBigException is thrown to client.  
  1377.         </description>  
  1378.     </property>  
  1379.     <property>  
  1380.         <name>hbase.thrift.minWorkerThreads</name>  
  1381.         <value>16</value>  
  1382.         <description>The "core size" of the thread pool. New threads are  
  1383.             created on every  
  1384.             connection until this many threads are created.  
  1385.         </description>  
  1386.     </property>  
  1387.     <property>  
  1388.         <name>hbase.thrift.maxWorkerThreads</name>  
  1389.         <value>1000</value>  
  1390.         <description>The maximum size of the thread pool. When the pending  
  1391.             request queue  
  1392.             overflows, new threads are created until their number reaches this number.  
  1393.             After that, the server starts dropping connections.  
  1394.         </description>  
  1395.     </property>  
  1396.     <property>  
  1397.         <name>hbase.thrift.maxQueuedRequests</name>  
  1398.         <value>1000</value>  
  1399.         <description>The maximum number of pending Thrift connections waiting  
  1400.             in the queue. If  
  1401.             there are no idle threads in the pool, the server queues requests. Only  
  1402.             when the queue overflows, new threads are added, up to  
  1403.             hbase.thrift.maxQueuedRequests threads.  
  1404.         </description>  
  1405.     </property>  
  1406.     <property>  
  1407.         <name>hbase.thrift.htablepool.size.max</name>  
  1408.         <value>1000</value>  
  1409.         <description>The upper bound for the table pool used in the Thrift  
  1410.             gateways server.  
  1411.             Since this is per table name, we assume a single table and so with 1000  
  1412.             default  
  1413.             worker threads max this is set to a matching number. For other workloads  
  1414.             this number  
  1415.             can be adjusted as needed.  
  1416.         </description>  
  1417.     </property>  
  1418.     <property>  
  1419.         <name>hbase.regionserver.thrift.framed</name>  
  1420.         <value>false</value>  
  1421.         <description>Use Thrift TFramedTransport on the server side.  
  1422.             This is the recommended transport for thrift servers and requires a  
  1423.             similar setting  
  1424.             on the client side. Changing this to false will select the default  
  1425.             transport,  
  1426.             vulnerable to DoS when malformed requests are issued due to THRIFT-601.  
  1427.         </description>  
  1428.     </property>  
  1429.     <property>  
  1430.         <name>hbase.regionserver.thrift.framed.max_frame_size_in_mb</name>  
  1431.         <value>2</value>  
  1432.         <description>Default frame size when using framed transport  
  1433.         </description>  
  1434.     </property>  
  1435.     <property>  
  1436.         <name>hbase.regionserver.thrift.compact</name>  
  1437.         <value>false</value>  
  1438.         <description>Use Thrift TCompactProtocol binary serialization  
  1439.             protocol.</description>  
  1440.     </property>  
  1441.     <property>  
  1442.         <name>hbase.rootdir.perms</name>  
  1443.         <value>700</value>  
  1444.         <description>FS Permissions for the root directory in a  
  1445.             secure(kerberos) setup.  
  1446.             When master starts, it creates the rootdir with this permissions or sets  
  1447.             the permissions  
  1448.             if it does not match.  
  1449.         </description>  
  1450.     </property>  
  1451.     <property>  
  1452.         <name>hbase.data.umask.enable</name>  
  1453.         <value>false</value>  
  1454.         <description>Enable, if true, that file permissions should be assigned  
  1455.             to the files written by the regionserver  
  1456.         </description>  
  1457.     </property>  
  1458.     <property>  
  1459.         <name>hbase.data.umask</name>  
  1460.         <value>000</value>  
  1461.         <description>File permissions that should be used to write data  
  1462.             files when hbase.data.umask.enable is true  
  1463.         </description>  
  1464.     </property>  
  1465.     <property>  
  1466.         <name>hbase.metrics.showTableName</name>  
  1467.         <value>true</value>  
  1468.         <description>Whether to include the prefix "tbl.tablename" in  
  1469.             per-column family metrics.  
  1470.             If true, for each metric M, per-cf metrics will be reported for  
  1471.             tbl.T.cf.CF.M, if false,  
  1472.             per-cf metrics will be aggregated by column-family across tables, and  
  1473.             reported for cf.CF.M.  
  1474.             In both cases, the aggregated metric M across tables and cfs will be  
  1475.             reported.  
  1476.         </description>  
  1477.     </property>  
  1478.     <property>  
  1479.         <name>hbase.metrics.exposeOperationTimes</name>  
  1480.         <value>true</value>  
  1481.         <description>Whether to report metrics about time taken performing an  
  1482.             operation on the region server. Get, Put, Delete, Increment, and  
  1483.             Append can all  
  1484.             have their times exposed through Hadoop metrics per CF and per region.  
  1485.         </description>  
  1486.     </property>  
  1487.     <!-- 允许快照被使用 -->  
  1488.     -->  
  1489.     <property>  
  1490.         <name>hbase.snapshot.enabled</name>  
  1491.         <value>true</value>  
  1492.         <description>Set to true to allow snapshots to be taken / restored /  
  1493.             cloned.</description>  
  1494.     </property>  
  1495.     <!-- 在hbase重启的时候，如果重启失败了，则使用快照代替，同时成功后删除快照 -->  
  1496.     <property>  
  1497.         <name>hbase.snapshot.restore.take.failsafe.snapshot</name>  
  1498.         <value>true</value>  
  1499.         <description>Set to true to take a snapshot before the restore  
  1500.             operation.  
  1501.             The snapshot taken will be used in case of failure, to restore the  
  1502.             previous state.  
  1503.             At the end of the restore operation this snapshot will be deleted  
  1504.         </description>  
  1505.     </property>  
  1506.     <property>  
  1507.         <name>hbase.snapshot.restore.failsafe.name</name>  
  1508.         <value>hbase-failsafe-{snapshot.name}-{restore.timestamp}</value>  
  1509.         <description>Name of the failsafe snapshot taken by the restore  
  1510.             operation.  
  1511.             You can use the {snapshot.name}, {table.name} and {restore.timestamp}  
  1512.             variables  
  1513.             to create a name based on what you are restoring.  
  1514.         </description>  
  1515.     </property>  
  1516.     <!-- hbase.server.compactchecker.interval.multiplier * hbase.server.thread.wakefrequency   
  1517.         后台线程每隔多久定期检查是否需要执行compaction -->  
  1518.     <property>  
  1519.         <name>hbase.server.compactchecker.interval.multiplier</name>  
  1520.         <value>1000</value>  
  1521.         <description>The number that determines how often we scan to see if  
  1522.             compaction is necessary.  
  1523.             Normally, compactions are done after some events (such as memstore flush), but  
  1524.             if  
  1525.             region didn't receive a lot of writes for some time, or due to different  
  1526.             compaction  
  1527.             policies, it may be necessary to check it periodically. The interval between  
  1528.             checks is  
  1529.             hbase.server.compactchecker.interval.multiplier multiplied by  
  1530.             hbase.server.thread.wakefrequency.  
  1531.         </description>  
  1532.     </property>  
  1533.     <property>  
  1534.         <name>hbase.lease.recovery.timeout</name>  
  1535.         <value>900000</value>  
  1536.         <description>How long we wait on dfs lease recovery in total before  
  1537.             giving up.</description>  
  1538.     </property>  
  1539.     <property>  
  1540.         <name>hbase.lease.recovery.dfs.timeout</name>  
  1541.         <value>64000</value>  
  1542.         <description>How long between dfs recover lease invocations. Should be  
  1543.             larger than the sum of  
  1544.             the time it takes for the namenode to issue a block recovery command as  
  1545.             part of  
  1546.             datanode; dfs.heartbeat.interval and the time it takes for the primary  
  1547.             datanode, performing block recovery to timeout on a dead datanode;  
  1548.             usually  
  1549.             dfs.client.socket-timeout. See the end of HBASE-8389 for more.  
  1550.         </description>  
  1551.     </property>  
  1552.     <!-- hbase colume最大的版本数 -->  
  1553.     <property>  
  1554.         <name>hbase.column.max.version</name>  
  1555.         <value>1</value>  
  1556.         <description>New column family descriptors will use this value as the  
  1557.             default number of versions  
  1558.             to keep.  
  1559.         </description>  
  1560.     </property>  
  1561.     <property>  
  1562.         <name>hbase.dfs.client.read.shortcircuit.buffer.size</name>  
  1563.         <value>131072</value>  
  1564.         <description>If the DFSClient configuration  
  1565.             dfs.client.read.shortcircuit.buffer.size is unset, we will  
  1566.             use what is configured here as the short circuit read default  
  1567.             direct byte buffer size. DFSClient native default is 1MB; HBase  
  1568.             keeps its HDFS files open so number of file blocks * 1MB soon  
  1569.             starts to add up and threaten OOME because of a shortage of  
  1570.             direct memory. So, we set it down from the default. Make  
  1571.             it > the default hbase block size set in the HColumnDescriptor  
  1572.             which is usually 64k.  
  1573.         </description>  
  1574.     </property>  
  1575.     <property>  
  1576.         <name>hbase.regionserver.checksum.verify</name>  
  1577.         <value>true</value>  
  1578.         <description>  
  1579.             If set to true (the default), HBase verifies the checksums for hfile  
  1580.             blocks. HBase writes checksums inline with the data when it writes  
  1581.             out  
  1582.             hfiles. HDFS (as of this writing) writes checksums to a separate file  
  1583.             than the data file necessitating extra seeks. Setting this flag saves  
  1584.             some on i/o. Checksum verification by HDFS will be internally  
  1585.             disabled  
  1586.             on hfile streams when this flag is set. If the hbase-checksum  
  1587.             verification  
  1588.             fails, we will switch back to using HDFS checksums (so do not disable HDFS  
  1589.             checksums! And besides this feature applies to hfiles only, not to  
  1590.             WALs).  
  1591.             If this parameter is set to false, then hbase will not verify any  
  1592.             checksums,  
  1593.             instead it will depend on checksum verification being done in the HDFS  
  1594.             client.  
  1595.         </description>  
  1596.     </property>  
  1597.     <property>  
  1598.         <name>hbase.hstore.bytes.per.checksum</name>  
  1599.         <value>16384</value>  
  1600.         <description>  
  1601.             Number of bytes in a newly created checksum chunk for HBase-level  
  1602.             checksums in hfile blocks.  
  1603.         </description>  
  1604.     </property>  
  1605.     <property>  
  1606.         <name>hbase.hstore.checksum.algorithm</name>  
  1607.         <value>CRC32C</value>  
  1608.         <description>  
  1609.             Name of an algorithm that is used to compute checksums. Possible values  
  1610.             are NULL, CRC32, CRC32C.  
  1611.         </description>  
  1612.     </property>  
  1613.     <!-- hbase客户端scan操作的时候，每次远程调用返回的最大字节数，默认是2M, 用来限制client从HRegionServer取到的bytes总数，bytes总数通过row的KeyValue计算得出 -->  
  1614.     <property>  
  1615.         <name>hbase.client.scanner.max.result.size</name>  
  1616.         <value>2097152</value>  
  1617.         <description>Maximum number of bytes returned when calling a scanner's  
  1618.             next method.  
  1619.             Note that when a single row is larger than this limit the row is still  
  1620.             returned completely.  
  1621.             The default value is 2MB, which is good for 1ge networks.  
  1622.             With faster and/or high latency networks this value should be increased.  
  1623.         </description>  
  1624.     </property>  
  1625.     <!-- hbase服务端对scan请求返回的结果大小做限制 -->  
  1626.     <property>  
  1627.         <name>hbase.server.scanner.max.result.size</name>  
  1628.         <value>104857600</value>  
  1629.         <description>Maximum number of bytes returned when calling a scanner's  
  1630.             next method.  
  1631.             Note that when a single row is larger than this limit the row is still  
  1632.             returned completely.  
  1633.             The default value is 100MB.  
  1634.             This is a safety setting to protect the server from OOM situations.  
  1635.         </description>  
  1636.     </property>  
  1637.     <property>  
  1638.         <name>hbase.status.published</name>  
  1639.         <value>false</value>  
  1640.         <description>  
  1641.             This setting activates the publication by the master of the status of the  
  1642.             region server.  
  1643.             When a region server dies and its recovery starts, the master will push  
  1644.             this information  
  1645.             to the client application, to let them cut the connection immediately  
  1646.             instead of waiting  
  1647.             for a timeout.  
  1648.         </description>  
  1649.     </property>  
  1650.     <property>  
  1651.         <name>hbase.status.publisher.class</name>  
  1652.         <value>org.apache.hadoop.hbase.master.ClusterStatusPublisher$MulticastPublisher  
  1653.         </value>  
  1654.         <description>  
  1655.             Implementation of the status publication with a multicast message.  
  1656.         </description>  
  1657.     </property>  
  1658.     <property>  
  1659.         <name>hbase.status.listener.class</name>  
  1660.         <value>org.apache.hadoop.hbase.client.ClusterStatusListener$MulticastListener  
  1661.         </value>  
  1662.         <description>  
  1663.             Implementation of the status listener with a multicast message.  
  1664.         </description>  
  1665.     </property>  
  1666.     <property>  
  1667.         <name>hbase.status.multicast.address.ip</name>  
  1668.         <value>226.1.1.3</value>  
  1669.         <description>  
  1670.             Multicast address to use for the status publication by multicast.  
  1671.         </description>  
  1672.     </property>  
  1673.     <property>  
  1674.         <name>hbase.status.multicast.address.port</name>  
  1675.         <value>16100</value>  
  1676.         <description>  
  1677.             Multicast port to use for the status publication by multicast.  
  1678.         </description>  
  1679.     </property>  
  1681.     <property>  
  1682.         <name>hbase.dynamic.jars.dir</name>  
  1683.         <value>${hbase.rootdir}/lib</value>  
  1684.         <description>  
  1685.             The directory from which the custom filter/co-processor jars can be  
  1686.             loaded  
  1687.             dynamically by the region server without the need to restart. However,  
  1688.             an already loaded filter/co-processor class would not be un-loaded. See  
  1689.             HBASE-1936 for more details.  
  1690.         </description>  
  1691.     </property>  
  1692.     <property>  
  1693.         <name>hbase.security.authentication</name>  
  1694.         <value>simple</value>  
  1695.         <description>  
  1696.             Controls whether or not secure authentication is enabled for HBase.  
  1697.             Possible values are 'simple' (no authentication), and 'kerberos'.  
  1698.         </description>  
  1699.     </property>  
  1700.     <property>  
  1701.         <name>hbase.rest.filter.classes</name>  
  1702.         <value>org.apache.hadoop.hbase.rest.filter.GzipFilter</value>  
  1703.         <description>  
  1704.             Servlet filters for REST service.  
  1705.         </description>  
  1706.     </property>  
  1707.     <property>  
  1708.         <name>hbase.master.loadbalancer.class</name>  
  1709.         <value>org.apache.hadoop.hbase.master.balancer.StochasticLoadBalancer  
  1710.         </value>  
  1711.         <description>  
  1712.             Class used to execute the regions balancing when the period occurs.  
  1713.             See the class comment for more on how it works  
  1714.             http://hbase.apache.org/devapidocs/org/apache/hadoop/hbase/master/balancer/StochasticLoadBalancer.html  
  1715.             It replaces the DefaultLoadBalancer as the default (since renamed  
  1716.             as the SimpleLoadBalancer).  
  1717.         </description>  
  1718.     </property>  
  1719.     <property>  
  1720.         <name>hbase.security.exec.permission.checks</name>  
  1721.         <value>false</value>  
  1722.         <description>  
  1723.             If this setting is enabled and ACL based access control is active (the  
  1724.             AccessController coprocessor is installed either as a system  
  1725.             coprocessor  
  1726.             or on a table as a table coprocessor) then you must grant all relevant  
  1727.             users EXEC privilege if they require the ability to execute  
  1728.             coprocessor  
  1729.             endpoint calls. EXEC privilege, like any other permission, can be  
  1730.             granted globally to a user, or to a user on a per table or per namespace  
  1731.             basis. For more information on coprocessor endpoints, see the  
  1732.             coprocessor  
  1733.             section of the HBase online manual. For more information on granting or  
  1734.             revoking permissions using the AccessController, see the security  
  1735.             section of the HBase online manual.  
  1736.         </description>  
  1737.     </property>  
  1738.     <property>  
  1739.         <name>hbase.procedure.regionserver.classes</name>  
  1740.         <value></value>  
  1741.         <description>A comma-separated list of  
  1742.             org.apache.hadoop.hbase.procedure.RegionServerProcedureManager  
  1743.             procedure managers that are  
  1744.             loaded by default on the active HRegionServer process. The lifecycle  
  1745.             methods (init/start/stop)  
  1746.             will be called by the active HRegionServer process to perform the  
  1747.             specific globally barriered  
  1748.             procedure. After implementing your own RegionServerProcedureManager, just put  
  1749.             it in  
  1750.             HBase's classpath and add the fully qualified class name here.  
  1751.         </description>  
  1752.     </property>  
  1753.     <property>  
  1754.         <name>hbase.procedure.master.classes</name>  
  1755.         <value></value>  
  1756.         <description>A comma-separated list of  
  1757.             org.apache.hadoop.hbase.procedure.MasterProcedureManager procedure  
  1758.             managers that are  
  1759.             loaded by default on the active HMaster process. A procedure is identified  
  1760.             by its signature and  
  1761.             users can use the signature and an instant name to trigger an execution of  
  1762.             a globally barriered  
  1763.             procedure. After implementing your own MasterProcedureManager, just put it in  
  1764.             HBase's classpath  
  1765.             and add the fully qualified class name here.  
  1766.         </description>  
  1767.     </property>  
  1768.     <property>  
  1769.         <name>hbase.coordinated.state.manager.class</name>  
  1770.         <value>org.apache.hadoop.hbase.coordination.ZkCoordinatedStateManager  
  1771.         </value>  
  1772.         <description>Fully qualified name of class implementing coordinated  
  1773.             state manager.</description>  
  1774.     </property>  
  1775.     <property>  
  1776.         <name>hbase.regionserver.storefile.refresh.period</name>  
  1777.         <value>0</value>  
  1778.         <description>  
  1779.             The period (in milliseconds) for refreshing the store files for the  
  1780.             secondary regions. 0  
  1781.             means this feature is disabled. Secondary regions sees new files (from  
  1782.             flushes and  
  1783.             compactions) from primary once the secondary region refreshes the list of files  
  1784.             in the  
  1785.             region (there is no notification mechanism). But too frequent refreshes  
  1786.             might cause  
  1787.             extra Namenode pressure. If the files cannot be refreshed for longer than  
  1788.             HFile TTL  
  1789.             (hbase.master.hfilecleaner.ttl) the requests are rejected. Configuring HFile TTL to a larger  
  1790.             value is also recommended with this setting.  
  1791.         </description>  
  1792.     </property>  
  1793.     <property>  
  1794.         <name>hbase.region.replica.replication.enabled</name>  
  1795.         <value>false</value>  
  1796.         <description>  
  1797.             Whether asynchronous WAL replication to the secondary region replicas is  
  1798.             enabled or not.  
  1799.             If this is enabled, a replication peer named  
  1800.             "region_replica_replication" will be created  
  1801.             which will tail the logs and replicate the mutatations to region replicas  
  1802.             for tables that  
  1803.             have region replication > 1. If this is enabled once, disabling this  
  1804.             replication also  
  1805.             requires disabling the replication peer using shell or ReplicationAdmin java  
  1806.             class.  
  1807.             Replication to secondary region replicas works over standard inter-cluster  
  1808.             replication.  
  1809.             So replication, if disabled explicitly, also has to be enabled by  
  1810.             setting "hbase.replication"  
  1811.             to true for this feature to work.  
  1812.         </description>  
  1813.     </property>  
  1814.     <property>  
  1815.         <name>hbase.http.filter.initializers</name>  
  1816.         <value>org.apache.hadoop.hbase.http.lib.StaticUserWebFilter</value>  
  1817.         <description>  
  1818.             A comma separated list of class names. Each class in the list must  
  1819.             extend  
  1820.             org.apache.hadoop.hbase.http.FilterInitializer. The corresponding Filter will  
  1821.             be initialized. Then, the Filter will be applied to all user facing jsp  
  1822.             and servlet web pages.  
  1823.             The ordering of the list defines the ordering of the filters.  
  1824.             The default StaticUserWebFilter add a user principal as defined by the  
  1825.             hbase.http.staticuser.user property.  
  1826.         </description>  
  1827.     </property>  
  1828.     <property>  
  1829.         <name>hbase.security.visibility.mutations.checkauths</name>  
  1830.         <value>false</value>  
  1831.         <description>  
  1832.             This property if enabled, will check whether the labels in the visibility  
  1833.             expression are associated  
  1834.             with the user issuing the mutation  
  1835.         </description>  
  1836.     </property>  
  1837.     <property>  
  1838.         <name>hbase.http.max.threads</name>  
  1839.         <value>10</value>  
  1840.         <description>  
  1841.             The maximum number of threads that the HTTP Server will create in its  
  1842.             ThreadPool.  
  1843.         </description>  
  1844.     </property>  
  1845.     <property>  
  1846.         <name>hbase.replication.rpc.codec</name>  
  1847.         <value>org.apache.hadoop.hbase.codec.KeyValueCodecWithTags</value>  
  1848.         <description>  
  1849.             The codec that is to be used when replication is enabled so that  
  1850.             the tags are also replicated. This is used along with HFileV3 which  
  1851.             supports tags in them. If tags are not used or if the hfile version  
  1852.             used  
  1853.             is HFileV2 then KeyValueCodec can be used as the replication codec.  
  1854.             Note that  
  1855.             using KeyValueCodecWithTags for replication when there are no tags causes  
  1856.             no harm.  
  1857.         </description>  
  1858.     </property>  
  1859.     <property>  
  1860.         <name>hbase.replication.source.maxthreads</name>  
  1861.         <value>10</value>  
  1862.         <description>  
  1863.             The maximum number of threads any replication source will use for  
  1864.             shipping edits to the sinks in parallel. This also limits the number  
  1865.             of  
  1866.             chunks each replication batch is broken into.  
  1867.             Larger values can improve the replication throughput between the master and  
  1868.             slave clusters. The default of 10 will rarely need to be changed.  
  1869.         </description>  
  1870.     </property>  
  1871.     <!-- Static Web User Filter properties. -->  
  1872.     <property>  
  1873.         <description>  
  1874.             The user name to filter as, on static web filters  
  1875.             while rendering content. An example use is the HDFS  
  1876.             web UI (user to be used for browsing files).  
  1877.         </description>  
  1878.         <name>hbase.http.staticuser.user</name>  
  1879.         <value>dr.stack</value>  
  1880.     </property>  
  1881.     <property>  
  1882.         <name>hbase.master.normalizer.class</name>  
  1883.         <value>org.apache.hadoop.hbase.master.normalizer.SimpleRegionNormalizer  
  1884.         </value>  
  1885.         <description>  
  1886.             Class used to execute the region normalization when the period occurs.  
  1887.             See the class comment for more on how it works  
  1888.             http://hbase.apache.org/devapidocs/org/apache/hadoop/hbase/master/normalizer/SimpleRegionNormalizer.html  
  1889.         </description>  
  1890.     </property>  
  1891.     <property>  
  1892.         <name>hbase.regionserver.handler.abort.on.error.percent</name>  
  1893.         <value>0.5</value>  
  1894.         <description>The percent of region server RPC threads failed to abort  
  1895.             RS.  
  1896.             -1 Disable aborting; 0 Abort if even a single handler has died;  
  1897.             0.x Abort only when this percent of handlers have died;  
  1898.             1 Abort only all of the handers have died.  
  1899.         </description>  
  1900.     </property>  
  1901.     <property>  
  1902.         <name>hbase.snapshot.master.timeout.millis</name>  
  1903.         <value>300000</value>  
  1904.         <description>  
  1905.             Timeout for master for the snapshot procedure execution  
  1906.         </description>  
  1907.     </property>  
  1908.     <property>  
  1909.         <name>hbase.snapshot.region.timeout</name>  
  1910.         <value>300000</value>  
  1911.         <description>  
  1912.             Timeout for regionservers to keep threads in snapshot request pool waiting  
  1913.         </description>  
  1914.     </property>  
  1915. </configuration>