[HBASE-17238] Wrong in-memory hbase:meta location causing SSH failure

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.1.0
• Fix Version/s: 1.3.0, 1.2.5, 1.1.9
• Component/s: Region Assignment
• Hadoop Flags: Reviewed
• Assignee: Stephen Yuan Jiang
• Reporter: Stephen Yuan Jiang
• Created: 02/Dec/16 08:22
• Updated: 09/Nov/17 02:43
• Resolved: 28/Dec/16 22:18

Description

In HBase 1.x, if HMaster#assignMeta() assigns a non-DEFAULT_REPLICA_ID hbase:meta region, it would wrongly update the DEFAULT_REPLICA_ID hbase:meta region in-memory. This caused the in-memory region state has wrong RS information for default replica hbase:meta region. One of the problem we saw is a wrong type of SSH could be chosen and causing problems.

void assignMeta(MonitoredTask status, Set<ServerName> previouslyFailedMetaRSs, int replicaId)
      throws InterruptedException, IOException, KeeperException {
    // Work on meta region
    ...
    if (replicaId == HRegionInfo.DEFAULT_REPLICA_ID) {
      status.setStatus("Assigning hbase:meta region");
    } else {
      status.setStatus("Assigning hbase:meta region, replicaId " + replicaId);
    }
    // Get current meta state from zk.
    RegionStates regionStates = assignmentManager.getRegionStates();
    RegionState metaState = MetaTableLocator.getMetaRegionState(getZooKeeper(), replicaId);
    HRegionInfo hri = RegionReplicaUtil.getRegionInfoForReplica(HRegionInfo.FIRST_META_REGIONINFO,
        replicaId);
    ServerName currentMetaServer = metaState.getServerName();
    ...
    boolean rit = this.assignmentManager.
      processRegionInTransitionAndBlockUntilAssigned(hri);
    boolean metaRegionLocation = metaTableLocator.verifyMetaRegionLocation(
      this.getConnection(), this.getZooKeeper(), timeout, replicaId);
    ...
        } else {
      // Region already assigned. We didn't assign it. Add to in-memory state.
      regionStates.updateRegionState(
        HRegionInfo.FIRST_META_REGIONINFO, State.OPEN, currentMetaServer); <<--- Wrong region to update -->>
      this.assignmentManager.regionOnline(
        HRegionInfo.FIRST_META_REGIONINFO, currentMetaServer); <<--- Wrong region to update -->>
    }
    ...

Here is the problem scenario:
Step 1: master failovers (or starts could have the same issue) and find default replica of hbase:meta is in rs1.

2016-11-26 00:06:36,590 INFO org.apache.hadoop.hbase.master.ServerManager: AssignmentManager hasn't finished failover cleanup; waiting
2016-11-26 00:06:36,591 INFO org.apache.hadoop.hbase.master.HMaster: hbase:meta with replicaId 0 assigned=0, rit=false, location=rs1,60200,1480103147220

Step 2: master finds that replica 1 of hbase:meta is unassigned, therefore, HMaster#assignMeta() is called and assign the replica 1 region to rs2, but at the end, it wrongly updates the in-memory state of default replica to rs2

2016-11-26 00:08:21,741 DEBUG org.apache.hadoop.hbase.master.RegionStates: Onlined 1588230740 on rs2,60200,1480102993815 {ENCODED => 1588230740, NAME => 'hbase:meta,,1', STARTKEY => '', ENDKEY => ''}
2016-11-26 00:08:21,741 INFO org.apache.hadoop.hbase.master.RegionStates: Offlined 1588230740 from rs1,60200,1480103147220
2016-11-26 00:08:21,741 INFO org.apache.hadoop.hbase.master.HMaster: hbase:meta with replicaId 1 assigned=0, rit=false, location=rs2,60200,1480102993815

Step 3: now rs1 is down, master needs to choose which SSH to call (MetaServerShutdownHandler or normal ServerShutdownHandler) - in this case, MetaServerShutdownHandler should be chosen; however, due to wrong in-memory location, normal ServerShutdownHandler was chosen:

2016-11-26 00:08:33,995 INFO org.apache.hadoop.hbase.zookeeper.RegionServerTracker: RegionServer ephemeral node deleted, processing expiration [rs1,60200,1480103147220]
2016-11-26 00:08:33,998 DEBUG org.apache.hadoop.hbase.master.AssignmentManager: based on AM, current region=hbase:meta,,1.1588230740 is on server=rs2,60200,1480102993815 server being checked: rs1,60200,1480103147220
2016-11-26 00:08:34,001 DEBUG org.apache.hadoop.hbase.master.ServerManager: Added=rs1,60200,1480103147220 to dead servers, submitted shutdown handler to be executed meta=false

Step 4: Wrong SSH was chosen. Due to accessing hbase:meta failure, the SSH failed after retries. Now the dead server was not processed; regions in that server remains un-usable (We have a solution that resolve this issue, but is undesirable: restarting master to cleanup the wrong in-memory state would make this problem go away):

2016-11-26 00:11:46,550 INFO org.apache.hadoop.hbase.master.handler.ServerShutdownHandler: Received exception accessing hbase:meta during server shutdown of rs1,60200,1480103147220, retrying hbase:meta read
org.apache.hadoop.hbase.client.RetriesExhaustedException: Failed after attempts=351, exceptions:
Sat Nov 26 00:11:46 EST 2016, null, java.net.SocketTimeoutException: callTimeout=60000, callDuration=68188: row '' on table 'hbase:meta' at region=hbase:meta,,1.1588230740, hostname=rs1,60200,1480103147220, seqNum=0

	at org.apache.hadoop.hbase.client.RpcRetryingCallerWithReadReplicas.throwEnrichedException(RpcRetryingCallerWithReadReplicas.java:271)
	at org.apache.hadoop.hbase.client.ScannerCallableWithReplicas.call(ScannerCallableWithReplicas.java:203)
	at org.apache.hadoop.hbase.client.ScannerCallableWithReplicas.call(ScannerCallableWithReplicas.java:60)
	at org.apache.hadoop.hbase.client.RpcRetryingCaller.callWithoutRetries(RpcRetryingCaller.java:200)
	at org.apache.hadoop.hbase.client.ClientScanner.call(ClientScanner.java:320)
	at org.apache.hadoop.hbase.client.ClientScanner.nextScanner(ClientScanner.java:295)
	at org.apache.hadoop.hbase.client.ClientScanner.initializeScannerInConstruction(ClientScanner.java:160)
	at org.apache.hadoop.hbase.client.ClientScanner.<init>(ClientScanner.java:155)
	at org.apache.hadoop.hbase.client.HTable.getScanner(HTable.java:794)
	at org.apache.hadoop.hbase.MetaTableAccessor.fullScan(MetaTableAccessor.java:602)
	at org.apache.hadoop.hbase.MetaTableAccessor.fullScan(MetaTableAccessor.java:156)
	at org.apache.hadoop.hbase.MetaTableAccessor.getServerUserRegions(MetaTableAccessor.java:555)
	at org.apache.hadoop.hbase.master.handler.ServerShutdownHandler.process(ServerShutdownHandler.java:177)
	at org.apache.hadoop.hbase.executor.EventHandler.run(EventHandler.java:129)
	at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
	at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
	at java.lang.Thread.run(Thread.java:745)
Caused by: java.net.SocketTimeoutException: callTimeout=60000, callDuration=68188: row '' on table 'hbase:meta' at region=hbase:meta,,1.1588230740, hostname=rs1,60200,1480103147220, seqNum=0
	at org.apache.hadoop.hbase.client.RpcRetryingCaller.callWithRetries(RpcRetryingCaller.java:159)
	at org.apache.hadoop.hbase.client.ResultBoundedCompletionService$QueueingFuture.run(ResultBoundedCompletionService.java:65)
	... 3 more
Caused by: java.io.IOException: Call to rs1/123.45.67.890:60200 failed on local exception: org.apache.hadoop.hbase.ipc.FailedServerException: This server is in the failed servers list: rs1/123.45.67.890:60200
	at org.apache.hadoop.hbase.ipc.RpcClientImpl.wrapException(RpcClientImpl.java:1261)
	at org.apache.hadoop.hbase.ipc.RpcClientImpl.call(RpcClientImpl.java:1229)
	at org.apache.hadoop.hbase.ipc.AbstractRpcClient.callBlockingMethod(AbstractRpcClient.java:213)
	at org.apache.hadoop.hbase.ipc.AbstractRpcClient$BlockingRpcChannelImplementation.callBlockingMethod(AbstractRpcClient.java:287)
	at org.apache.hadoop.hbase.protobuf.generated.ClientProtos$ClientService$BlockingStub.scan(ClientProtos.java:32651)
	at org.apache.hadoop.hbase.client.ScannerCallable.openScanner(ScannerCallable.java:372)
	at org.apache.hadoop.hbase.client.ScannerCallable.call(ScannerCallable.java:199)
	at org.apache.hadoop.hbase.client.ScannerCallable.call(ScannerCallable.java:62)
	at org.apache.hadoop.hbase.client.RpcRetryingCaller.callWithoutRetries(RpcRetryingCaller.java:200)
	at org.apache.hadoop.hbase.client.ScannerCallableWithReplicas$RetryingRPC.call(ScannerCallableWithReplicas.java:356)
	at org.apache.hadoop.hbase.client.ScannerCallableWithReplicas$RetryingRPC.call(ScannerCallableWithReplicas.java:330)
	at org.apache.hadoop.hbase.client.RpcRetryingCaller.callWithRetries(RpcRetryingCaller.java:126)
	... 4 more
Caused by: org.apache.hadoop.hbase.ipc.FailedServerException: This server is in the failed servers list: rs1/123.45.67.890:60200
	at org.apache.hadoop.hbase.ipc.RpcClientImpl$Connection.setupIOstreams(RpcClientImpl.java:701)
	at org.apache.hadoop.hbase.ipc.RpcClientImpl$Connection.writeRequest(RpcClientImpl.java:887)
	at org.apache.hadoop.hbase.ipc.RpcClientImpl$Connection.tracedWriteRequest(RpcClientImpl.java:856)
	at org.apache.hadoop.hbase.ipc.RpcClientImpl$Connection$CallSender.run(RpcClientImpl.java:266)

Issue Links

• Is contained by: HBASE-18223 Track the effort to improve/bug fix read replica feature
• relates to: HBASE-17597 TestMetaWithReplicas.testMetaTableReplicaAssignment is flaky