Rushabh Shah created HBASE-26963: ------------------------------------ Summary: ReplicationSource#removePeer hangs if we try to remove bad peer. Key: HBASE-26963 URL: https://issues.apache.org/jira/browse/HBASE-26963 Project: HBase Issue Type: Bug Affects Versions: 2.4.11 Reporter: Rushabh Shah
ReplicationSource#removePeer hangs if we try to remove bad peer. Steps to reproduce: 1. Set config replication.source.regionserver.abort to false so that it doesn't abort regionserver. 2. Add a dummy peer. 2. Remove that peer. RemovePeer call will hang indefinitely until the test times out. Attached a patch to reproduce the above behavior. I can see following threads in the stack trace: {noformat} "RS_REFRESH_PEER-regionserver/rushabh-ltmflld:0-0.replicationSource,dummypeer_1" #339 daemon prio=5 os_prio=31 tid=0x00007f8caa 44a800 nid=0x22107 waiting on condition [0x00007000107e5000] java.lang.Thread.State: TIMED_WAITING (sleeping) at java.lang.Thread.sleep(Native Method) at org.apache.hadoop.hbase.replication.regionserver.ReplicationSource.sleepForRetries(ReplicationSource.java:511) at org.apache.hadoop.hbase.replication.regionserver.ReplicationSource.initialize(ReplicationSource.java:577) at org.apache.hadoop.hbase.replication.regionserver.ReplicationSource.lambda$startup$4(ReplicationSource.java:633) at org.apache.hadoop.hbase.replication.regionserver.ReplicationSource$$Lambda$350/89698794.uncaughtException(Unknown Source) at java.lang.Thread.dispatchUncaughtException(Thread.java:1959) {noformat} {noformat} "RS_REFRESH_PEER-regionserver/rushabh-ltmflld:0-0" #338 daemon prio=5 os_prio=31 tid=0x00007f8ca82fa800 nid=0x22307 in Object.wait() [0x00007000106e2000] java.lang.Thread.State: TIMED_WAITING (on object monitor) at java.lang.Object.wait(Native Method) at java.lang.Thread.join(Thread.java:1260) - locked <0x0000000799975ea0> (a java.lang.Thread) at org.apache.hadoop.hbase.util.Threads.shutdown(Threads.java:106) at org.apache.hadoop.hbase.replication.regionserver.ReplicationSource.terminate(ReplicationSource.java:674) at org.apache.hadoop.hbase.replication.regionserver.ReplicationSource.terminate(ReplicationSource.java:657) at org.apache.hadoop.hbase.replication.regionserver.ReplicationSource.terminate(ReplicationSource.java:652) at org.apache.hadoop.hbase.replication.regionserver.ReplicationSource.terminate(ReplicationSource.java:647) at org.apache.hadoop.hbase.replication.regionserver.ReplicationSourceManager.removePeer(ReplicationSourceManager.java:330) at org.apache.hadoop.hbase.replication.regionserver.PeerProcedureHandlerImpl.removePeer(PeerProcedureHandlerImpl.java:56) at org.apache.hadoop.hbase.replication.regionserver.RefreshPeerCallable.call(RefreshPeerCallable.java:61) at org.apache.hadoop.hbase.replication.regionserver.RefreshPeerCallable.call(RefreshPeerCallable.java:35) at org.apache.hadoop.hbase.regionserver.handler.RSProcedureHandler.process(RSProcedureHandler.java:49) at org.apache.hadoop.hbase.executor.EventHandler.run(EventHandler.java:104) at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149) at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624) at java.lang.Thread.run(Thread.java:748) {noformat} {noformat} "Listener at localhost/55013" #20 daemon prio=5 os_prio=31 tid=0x00007f8caf95a000 nid=0x6703 waiting on condition [0x0000700002 544000] java.lang.Thread.State: TIMED_WAITING (sleeping) at java.lang.Thread.sleep(Native Method) at org.apache.hadoop.hbase.client.HBaseAdmin$ProcedureFuture.waitProcedureResult(HBaseAdmin.java:3442) at org.apache.hadoop.hbase.client.HBaseAdmin$ProcedureFuture.get(HBaseAdmin.java:3372) at org.apache.hadoop.hbase.util.FutureUtils.get(FutureUtils.java:182) at org.apache.hadoop.hbase.client.Admin.removeReplicationPeer(Admin.java:2861) at org.apache.hadoop.hbase.client.replication.TestBadReplicationPeer.cleanPeer(TestBadReplicationPeer.java:74) at org.apache.hadoop.hbase.client.replication.TestBadReplicationPeer.testWrongReplicationEndpoint(TestBadReplicationPeer.java:66) {noformat} The main thread "TestBadReplicationPeer.testWrongReplicationEndpoint" is waiting for Admin#removeReplicationPeer. The refreshPeer thread (PeerProcedureHandlerImpl#removePeer) responsible to terminate peer (#338) is waiting on ReplicationSource thread to be terminated. The ReplicateSource thread (#339) is in sleeping state. Notice that this thread's stack trace is in ReplicationSource#uncaughtException method. When we call ReplicationSourceManager#removePeer, we set sourceRunning flag to false, send an interrupt signal to ReplicationSource thread [here|https://github.com/apache/hbase/blob/branch-2.4/hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/ReplicationSource.java#L668-L674]. In this case ReplicationSource was waiting to read cluster id of the peer and it received an InterruptedException. {noformat} 2022-04-20 08:46:49,679 WARN [RS_REFRESH_PEER-regionserver/rushabh-ltmflld:0-0.replicationSource,dummypeer_1] zookeeper.ZKUtil(228): connection to cluster: dummypeer_1-0x100229efa200009, quorum=127.0.0.1:55599, baseZNode=/1 Unable to set watcher on znode (/1/hbaseid) java.lang.InterruptedException at java.lang.Object.wait(Native Method) at java.lang.Object.wait(Object.java:502) at org.apache.zookeeper.ClientCnxn.submitRequest(ClientCnxn.java:1529) at org.apache.zookeeper.ClientCnxn.submitRequest(ClientCnxn.java:1512) at org.apache.zookeeper.ZooKeeper.exists(ZooKeeper.java:2016) at org.apache.hadoop.hbase.zookeeper.RecoverableZooKeeper.exists(RecoverableZooKeeper.java:212) at org.apache.hadoop.hbase.zookeeper.ZKUtil.checkExists(ZKUtil.java:221) at org.apache.hadoop.hbase.zookeeper.ZKClusterId.readClusterIdZNode(ZKClusterId.java:65) at org.apache.hadoop.hbase.zookeeper.ZKClusterId.getUUIDForCluster(ZKClusterId.java:96) at org.apache.hadoop.hbase.replication.HBaseReplicationEndpoint.getPeerUUID(HBaseReplicationEndpoint.java:112) at org.apache.hadoop.hbase.replication.regionserver.ReplicationSource.initialize(ReplicationSource.java:571) at java.lang.Thread.run(Thread.java:748) {noformat} [ZKClusterId.readClusterIdZNode|https://github.com/apache/hbase/blob/branch-2.4/hbase-zookeeper/src/main/java/org/apache/hadoop/hbase/zookeeper/ZKClusterId.java#L69-L72] catches InterruptedException and returns null. ReplicationSource realizes that sourceRunning flag is set to false and it will throw IllegalStateException [here|https://github.com/apache/hbase/blob/branch-2.4/hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/ReplicationSource.java#L561-L565]. Then the control goes to [UncaughtExceptionHandler|https://github.com/apache/hbase/blob/branch-2.4/hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/ReplicationSource.java#L620-L640] and since abortOnError is set to false, it will go into infinite sleep causing the test to hang. -- This message was sent by Atlassian Jira (v8.20.7#820007)