hbase 修复 hbck


hbase 修复使用hbck 

 

新版本的 hbck 可以修复各种错误,修复选项是:     
(1)-fix,向下兼容用,被-fixAssignments替代     
(2)-fixAssignments,用于修复region assignments错误     
(3)-fixMeta,用于修复meta表的问题,前提是HDFS上面的region info信息有并且正确。     
(4)-fixHdfsHoles,修复region holes(空洞,某个区间没有region)问题     
(5)-fixHdfsOrphans,修复Orphan region(hdfs上面没有.regioninfo的region)     
(6)-fixHdfsOverlaps,修复region overlaps(区间重叠)问题     
(7)-fixVersionFile,修复缺失hbase.version文件的问题     
(8)-maxMerge <n> (n默认是5),当region有重叠是,需要合并region,一次合并的region数最大不超过这个值。     
(9)-sidelineBigOverlaps ,当修复region overlaps问题时,允许跟其他region重叠次数最多的一些region不参与(修复后,可以把没有参与的数据通过bulk load加载到相应的region)     
(10)-maxOverlapsToSideline <n> (n默认是2),当修复region overlaps问题时,一组里最多允许多少个region不参与     
由于选项较多,所以有两个简写的选项     
(11) -repair,相当于-fixAssignments -fixMeta -fixHdfsHoles -fixHdfsOrphans -fixHdfsOverlaps -fixVersionFile -sidelineBigOverlaps     
(12)-repairHoles,相当于-fixAssignments -fixMeta -fixHdfsHoles -fixHdfsOrphans     
     
     
     
新版本的 hbck     
(1)缺失hbase.version文件     
 加上选项 -fixVersionFile 解决     
(2)如果一个region即不在META表中,又不在hdfs上面,但是在regionserver的online region集合中     
 加上选项 -fixAssignments 解决     
(3)如果一个region在META表中,并且在regionserver的online region集合中,但是在hdfs上面没有     
 加上选项 -fixAssignments -fixMeta 解决,( -fixAssignments告诉regionserver close region),( -fixMeta删除META表中region的记录)     
(4)如果一个region在META表中没有记录,没有被regionserver服务,但是在hdfs上面有     
加上选项 -fixMeta -fixAssignments 解决,( -fixAssignments 用于assign region),( -fixMeta用于在META表中添加region的记录)     
(5)如果一个region在META表中没有记录,在hdfs上面有,被regionserver服务了     
加上选项 -fixMeta 解决,在META表中添加这个region的记录,先undeploy region,后assign     
(6)如果一个region在META表中有记录,但是在hdfs上面没有,并且没有被regionserver服务     
加上选项 -fixMeta 解决,删除META表中的记录     
(7)如果一个region在META表中有记录,在hdfs上面也有,table不是disabled的,但是这个region没有被服务     
加上选项 -fixAssignments 解决,assign这个region     
(8)如果一个region在META表中有记录,在hdfs上面也有,table是disabled的,但是这个region被某个regionserver服务了     
加上选项 -fixAssignments 解决,undeploy这个region     
(9)如果一个region在META表中有记录,在hdfs上面也有,table不是disabled的,但是这个region被多个regionserver服务了     
加上选项 -fixAssignments 解决,通知所有regionserver close region,然后assign region     
(10)如果一个region在META表中,在hdfs上面也有,也应该被服务,但是META表中记录的regionserver和实际所在的regionserver不相符     
加上选项 -fixAssignments 解决     
     
(11)region holes     
需要加上 -fixHdfsHoles ,创建一个新的空region,填补空洞,但是不assign 这个 region,也不在META表中添加这个region的相关信息     
(12)region在hdfs上面没有.regioninfo文件     
-fixHdfsOrphans 解决     
(13)region overlaps     
需要加上 -fixHdfsOverlaps     
     
     
说明:     
(1)修复region holes时,-fixHdfsHoles 选项只是创建了一个新的空region,填补上了这个区间,还需要加上-fixAssignments -fixMeta 来解决问题,
    -fixAssignments 用于assign region,
    -fixMeta用于在META表中添加region的记录,所以有了组合拳 -repairHoles 修复region holes,相当于-fixAssignments -fixMeta -fixHdfsHoles -fixHdfsOrphans     
(2) -fixAssignments,用于修复region没有assign、不应该assign、assign了多次的问题     
(3)-fixMeta,如果hdfs上面没有,那么从META表中删除相应的记录,如果hdfs上面有,在META表中添加上相应的记录信息     
(4)-repair 打开所有的修复选项,相当于-fixAssignments -fixMeta
-fixHdfsHoles -fixHdfsOrphans -fixHdfsOverlaps -fixVersionFile -sidelineBigOverlaps     
     
新版本的hbck从(1)hdfs目录(2)META(3)RegionServer这三处获得region的Table和Region的相关信息,根据这些信息判断并repair    

 

事例:

 

查看hbasemeta情况    
hbase hbck    
1.重新修复hbase meta表(根据hdfs上的regioninfo文件,生成meta表)    
  hbase hbck -fixMeta    
2.重新将hbase meta表分给regionserver(根据meta表,将meta表上的region分给regionservere)    
  hbase hbck -fixAssignments    

当出现漏洞 
hbase hbck -fixHdfsHoles  (新建一个region文件夹) 
hbase hbck -fixMeta (根据regioninfo生成meta表)
hbase hbck -fixAssignments  (分配region到regionserver上)

 

一、故障原因    
IP为10.191.135.3的服务器在2013年8月1日出现服务器重新启动的情况,导致此台服务器上的所有服务均停止。
从而造成NTP服务停止。当NTP服务停止后,导致HBase集群中大部分机器时钟和主机时间不一致,造成regionserver服务中止。
并在重新启动后,出现region的hole。需要对数据进行重新修复,以正常提供插入数据的服务。 二、恢复方式
1、集群50个regionserver,宕掉服务41个,namenode所在机器10.191.135.3不明重启(原因查找中)导致本机上的namenode、zookeeper、时间同步服务器服务挂掉。 2、重启hbase服务时,没能成功stop剩余的9个regionserver服务,进行了人为kill进程, 3、在hdfs上移走了hlog(避免启动时split log花费过多时间影响服务),然后重启hbase。发现10.191.135.30机器上的时间与时间同步服务器10.191.135.3不同步。
  手工同步后重启成功。hbase可以正常提供查询服务。
4、运行mapreduce put数据。抛出异常,数据无法正常插入; 5、执行/opt/hbase/bin/hbase hbck -fixAssignments,尝试重新分配region。结果显示hbase有空洞,即region之间数据不连续了; 6、通过上述操作可以定位是在regionserver服务宕掉的后重启的过程中丢了数据。需要进行空洞修复。然而hbase hbck命令总是只显示三条空洞。 7、通过编写的regionTest.jar工具进行进一步检测出空洞所在的regionname然后停掉hbase,进而进行region合并修复空洞; 8、合并的merge 操作需要先去.META.表里读取该region的信息,由于.META.表也在regionserver宕机过程中受到损坏,所以部分region的.META.信息没有,
  merge操作时就抛出空指针异常。因此只能将hdfs这些region进行移除,然后通过regionTest.jar 检测新的空洞所在的regionname,进行合并操作修复空洞;
9、关于region重叠,即regionname存在.META.表内,但是在hdfs上被错误的移出,并进行了region合并。这种情况下需要通过regionTest.jar检测重叠的regionname然后手动去.META.表删除,
  .META.表修改之后需要flush;
10、最后再次执行 hbase hbck 命令,hbase 所有表status ok。 三、相关命令及页面报错信息 1.手工同步时间命令
service ntpd stop
ntpdate -d 192.168.1.20
service ntpd start 2.org.apache.hadoop.hbase.client.RetriesExhaustedWithDetailsException: Failed 2 actions: WrongRegionException: 2 times, servers with issues: datanode10:60020, 
at org.apache.hadoop.hbase.client.HConnectionManager$HConnectionImplementation.processBatchCallback(HConnectionManager.java:1641)
at org.apache.hadoop.hbase.client.HConnectionManager$HConnectionImplementation.processBatch(HConnectionManager.java:1409)
at org.apache.hadoop.hbase.client.HTable.flushCommits(HTable.java:949)
at org.apache.hadoop.hbase.client.HTable.doPut(HTable.java:826)
at org.apache.hadoop.hbase.client.HTable.put(HTable.java:801)
at org.apache.hadoop.hbase.mapreduce.TableOutputFormat$TableRecordWriter.write(TableOutputFormat.java:123)
at org.apache.hadoop.hbase.mapreduce.TableOutputFormat$TableRecordWriter.write(TableOutputFormat.java:84)
at org.apache.hadoop.mapred.MapTask$NewDirectOutputCollector.write(MapTask.java:533)
at org.apache.hadoop.mapreduce.task.TaskInputOutputContextImpl.write(TaskInputOutputContextImpl.java:88)
at o 3.13/08/01 18:30:02 DEBUG util.HBaseFsck: There are 22093 region info entries
ERROR: There is a hole in the region chain between +8615923208069cmnet201303072132166264580 and +861592321. You need to create a new .regioninfo and region dir in hdfs to plug the hole.
ERROR: There is a hole in the region chain between +8618375993383cmwap20130512235639430 and +8618375998629cmnet201305040821436779670. You need to create a new .regioninfo and region dir in hdfs to plug the hole.
ERROR: There is a hole in the region chain between +8618725888080cmnet201212271719506311400 and +8618725889786cmnet201302131646431671140. You need to create a new .regioninfo and region dir in hdfs to plug the hole.
ERROR: Found inconsistency in table cqgprs
Summary:
 -ROOT- is okay.
 Number of regions: 1
 Deployed on: datanode14,60020,1375330955915
 .META. is okay.
 Number of regions: 1
 Deployed on: datanode21,60020,1375330955825
 cqgprs is okay.
 Number of regions: 22057
 Deployed on: datanode1,60020,1375330955761 datanode10,60020,1375330955748 datanode11,60020,1375330955736 datanode12,60020,1375330955993 datanode13,60020,1375330955951 datanode14,60020,1375330955915 datanode15,60020,1375330955882 datanode16,60020,1375330955892 datanode17,60020,1375330955864 datanode18,60020,1375330955703 datanode19,60020,1375330955910 datanode2,60020,1375330955751 datanode20,60020,1375330955849 datanode21,60020,1375330955825 datanode22,60020,1375334479752 datanode23,60020,1375330955835 datanode24,60020,1375330955932 datanode25,60020,1375330955856 datanode26,60020,1375330955807 datanode27,60020,1375330955882 datanode28,60020,1375330955785 datanode29,60020,1375330955799 datanode3,60020,1375330955778 datanode30,60020,1375330955748 datanode31,60020,1375330955877 datanode32,60020,1375330955763 datanode33,60020,1375330955755 datanode34,60020,1375330955713 datanode35,60020,1375330955768 datanode36,60020,1375330955896 datanode37,60020,1375330955884 datanode38,60020,1375330955918 datanode39,60020,1375330955881 datanode4,60020,1375330955826 datanode40,60020,1375330955770 datanode41,60020,1375330955824 datanode42,60020,1375449245386 datanode43,60020,1375330955880 datanode44,60020,1375330955902 datanode45,60020,1375330955881 datanode46,60020,1375330955841 datanode47,60020,1375330955790 datanode48,60020,1375330955848 datanode49,60020,1375330955849 datanode5,60020,1375330955880 datanode50,60020,1375330955802 datanode6,60020,1375330955753 datanode7,60020,1375330955890 datanode8,60020,1375330955967 datanode9,60020,1375330955948
 test1 is okay.
 Number of regions: 1
 Deployed on: datanode43,60020,1375330955880
 test2 is okay.
 Number of regions: 1
 Deployed on: datanode21,60020,1375330955825
35 inconsistencies detected.
Status: INCONSISTENT 4.hadoop jar regionTest.jar com.region.RegionReaderMain /hbase/cqgprs 检测cqgprs表里的空洞所在的regionname。 5.==================================
first endKey = +8615808059207cmnet201307102326567966800
second startKey = +8615808058578cmnet201212251545557984830

first regionNmae = cqgprs,+8615808058578cmnet201212251545557984830,1375241186209.0f8266ad7ac45be1fa7233e8ea7aeef9.
second regionNmae = cqgprs,+8615808058578cmnet201212251545557984830,1362778571889.3552d3db8166f421047525d6be39c22e.
==================================
first endKey = +8615808060140cmnet201303051801355846850
second startKey = +8615808059207cmnet201307102326567966800

first regionNmae = cqgprs,+8615808058578cmnet201212251545557984830,1362778571889.3552d3db8166f421047525d6be39c22e.
second regionNmae = cqgprs,+8615808059207cmnet201307102326567966800,1375241186209.09d489d3df513bc79bab09cec36d2bb4.
================================== 6.Usage: bin/hbase org.apache.hadoop.hbase.util.Merge [-Dfs.default.name=hdfs://nn:port] <table-name> <region-1> <region-2>

./hbase org.apache.hadoop.hbase.util.Merge -Dfs.defaultFS=hdfs://bdpha cqgprs cqgprs,+8615213741567cmnet201305251243290802280,1369877465524.3c13b460fae388b1b1a70650b66c5039. cqgprs,+8615213745577cmnet201302141725552206710,1369534940433.5de80f59071555029ac42287033a4863. & 7.13/08/01 22:24:02 WARN util.HBaseFsck: Naming new problem group: +8618225125357cmnet201212290358070667800
ERROR: (regions cqgprs,+8618225123516cmnet201304131404096748520,1375363774655.b3cf5cc752f4427a4e699270dff9839e. and cqgprs,+8618225125357cmnet201212290358070667800,1364421610707.7f7038bfbe2c0df0998a529686a3e1aa.) There is an overlap in the region chain.
13/08/01 22:24:02 WARN util.HBaseFsck: reached end of problem group: +8618225127504cmnet201302182135452100210
13/08/01 22:24:02 WARN util.HBaseFsck: Naming new problem group: +8618285642723cmnet201302031921019768070
ERROR: (regions cqgprs,+8618285277826cmnet201306170027424674330,1375363962312.9d1e93b22cec90fd75361fa65b1d20d2. and cqgprs,+8618285642723cmnet201302031921019768070,1360873307626.f631cd8c6acc5e711e651d13536abe94.) There is an overlap in the region chain.
13/08/01 22:24:02 WARN util.HBaseFsck: reached end of problem group: +8618286275556cmnet201212270713444340110
13/08/01 22:24:02 WARN util.HBaseFsck: Naming new problem group: +8618323968833cmnet201306010239025175240
ERROR: (regions cqgprs,+8618323967956cmnet201306091923411365860,1375364143678.665dba6a14ebc9971422b39e079b00ae. and cqgprs,+8618323968833cmnet201306010239025175240,1372821719159.6d2fecc1b3f9049bbca83d84231eb365.) There is an overlap in the region chain.
13/08/01 22:24:02 WARN util.HBaseFsck: reached end of problem group: +8618323992353cmnet201306012336364819810
ERROR: There is a hole in the region chain between +8618375993383cmwap20130512235639430 and +8618375998629cmnet201305040821436779670. You need to create a new .regioninfo and region dir in hdfs to plug the hole.
13/08/01 22:24:02 WARN util.HBaseFsck: Naming new problem group: +8618723686187cmnet201301191433522129820
ERROR: (regions cqgprs,+8618723683087cmnet201301300708363045080,1375364411992.4ee5787217c1da4895d95b3b92b8e3a2. and cqgprs,+8618723686187cmnet201301191433522129820,1362003066106.70b48899cc753a0036f11bb27d2194f9.) There is an overlap in the region chain.
13/08/01 22:24:02 WARN util.HBaseFsck: reached end of problem group: +8618723689138cmnet201301051742388948390
13/08/01 22:24:02 WARN util.HBaseFsck: Naming new problem group: +8618723711808cmnet201301031139206225900
ERROR: (regions cqgprs,+8618723710003cmnet201301250809235976320,1375364586329.40eed10648c9a43e3d5ce64e9d63fe00. and cqgprs,+8618723711808cmnet201301031139206225900,1361216401798.ebc442e02f5e784bce373538e06dd232.) There is an overlap in the region chain.
13/08/01 22:24:02 WARN util.HBaseFsck: reached end of problem group: +8618723714626cmnet201302122009459491970
ERROR: There is a hole in the region chain between +8618725888080cmnet201212271719506311400 and +8618725889786cmnet201302131646431671140. You need to create a new .regioninfo and region dir in hdfs to plug the hole. 8. delete '.META.','regionname','info:serverstartcode' delete '.META.','regionname','info:regionserver' delete '.META.','regionname','info:regioninfo' 9. flush '.META.'
major_compact '.META.'

 

posted @ 2017-05-27 15:23  新际航  阅读(3179)  评论(0编辑  收藏  举报