[
https://issues.apache.org/jira/browse/HBASE-25612?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17298843#comment-17298843
]
Rushabh Shah commented on HBASE-25612:
--------------------------------------
> ReplicationLogCleaner uses its own zk client session?
In master and branch-2,
[ReplicationLogCleaner|https://github.com/apache/hbase/blob/master/hbase-server/src/main/java/org/apache/hadoop/hbase/replication/master/ReplicationLogCleaner.java#L102-L107]
shares zk session with HMaster.
But in branch-1,
[ReplicationLogCleaner|https://github.com/apache/hbase/blob/branch-1/hbase-server/src/main/java/org/apache/hadoop/hbase/replication/master/ReplicationLogCleaner.java#L139]
creates its own ZKWatcher object.
[~anoop.hbase] We observed this behavior in cluster which is running hbase-1.3
version. Looks like this bug exists only in branch-1
> HMaster should abort if ReplicationLogCleaner is not able to delete oldWALs.
> ----------------------------------------------------------------------------
>
> Key: HBASE-25612
> URL: https://issues.apache.org/jira/browse/HBASE-25612
> Project: HBase
> Issue Type: Improvement
> Affects Versions: 1.6.0
> Reporter: Rushabh Shah
> Assignee: Rushabh Shah
> Priority: Major
>
> In our production cluster, we encountered an issue where the number of files
> within /hbase/oldWALs directory were growing exponentially from about 4000
> baseline to 150000 and growing at the rate of 333 files per minute.
> On further investigation we found that ReplicatonLogCleaner thread was
> getting aborted since it was not able to talk to zookeeper. Stack trace below
> {noformat}
> 2021-02-25 23:05:01,149 WARN [an-pool3-thread-1729] zookeeper.ZKUtil -
> replicationLogCleaner-0x3000002e05e0d8f,
> quorum=zookeeper-0:2181,zookeeper-1:2181,zookeeper-2:2181,zookeeper-3:2181,zookeeper-4:2181,
> baseZNode=/hbase Unable to get data of znode /hbase/replication/rs
> org.apache.zookeeper.KeeperException$SessionExpiredException: KeeperErrorCode
> = Session expired for /hbase/replication/rs
> at org.apache.zookeeper.KeeperException.create(KeeperException.java:130)
> at org.apache.zookeeper.KeeperException.create(KeeperException.java:54)
> at org.apache.zookeeper.ZooKeeper.getData(ZooKeeper.java:1229)
> at
> org.apache.hadoop.hbase.zookeeper.RecoverableZooKeeper.getData(RecoverableZooKeeper.java:374)
> at org.apache.hadoop.hbase.zookeeper.ZKUtil.getDataNoWatch(ZKUtil.java:713)
> at
> org.apache.hadoop.hbase.replication.ReplicationQueuesClientZKImpl.getQueuesZNodeCversion(ReplicationQueuesClientZKImpl.java:87)
> at
> org.apache.hadoop.hbase.replication.master.ReplicationLogCleaner.loadWALsFromQueues(ReplicationLogCleaner.java:99)
> at
> org.apache.hadoop.hbase.replication.master.ReplicationLogCleaner.getDeletableFiles(ReplicationLogCleaner.java:70)
> at
> org.apache.hadoop.hbase.master.cleaner.CleanerChore.checkAndDeleteFiles(CleanerChore.java:262)
> at
> org.apache.hadoop.hbase.master.cleaner.CleanerChore.access$200(CleanerChore.java:52)
> at
> org.apache.hadoop.hbase.master.cleaner.CleanerChore$3.act(CleanerChore.java:413)
> at
> org.apache.hadoop.hbase.master.cleaner.CleanerChore$3.act(CleanerChore.java:410)
> at
> org.apache.hadoop.hbase.master.cleaner.CleanerChore.deleteAction(CleanerChore.java:481)
> at
> org.apache.hadoop.hbase.master.cleaner.CleanerChore.traverseAndDelete(CleanerChore.java:410)
> at
> org.apache.hadoop.hbase.master.cleaner.CleanerChore.access$100(CleanerChore.java:52)
> at
> org.apache.hadoop.hbase.master.cleaner.CleanerChore$1.run(CleanerChore.java:220)
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
> at java.lang.Thread.run(Thread.java:748)
> 2021-02-25 23:05:01,149 WARN [an-pool3-thread-1729]
> master.ReplicationLogCleaner - ReplicationLogCleaner received abort,
> ignoring. Reason: Failed to get stat of replication rs node
> 2021-02-25 23:05:01,149 DEBUG [an-pool3-thread-1729]
> master.ReplicationLogCleaner -
> org.apache.zookeeper.KeeperException$SessionExpiredException: KeeperErrorCode
> = Session expired for /hbase/replication/rs
> 2021-02-25 23:05:01,150 WARN [an-pool3-thread-1729]
> master.ReplicationLogCleaner - Failed to read zookeeper, skipping checking
> deletable files
> {noformat}
>
> {quote} 2021-02-25 23:05:01,149 WARN [an-pool3-thread-1729]
> master.ReplicationLogCleaner - ReplicationLogCleaner received abort,
> ignoring. Reason: Failed to get stat of replication rs node
> {quote}
>
> This line is more scary where HMaster invoked Abortable but just ignored and
> HMaster was doing it business as usual.
> We have max files per directory configuration in namenode which is set to 1M
> in our clusters. If this directory reached that limit then that would have
> brought down the whole cluster.
> We shouldn't ignore Abortable and should crash the Hmaster if Abortable is
> invoked.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
