[HBASE-28058] HMaster snapshot file clean thread and the snapshot request handler thread encountered a deadlock

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.0.0-beta-1
• Fix Version/s: 2.6.0, 3.0.0-beta-1
• Component/s: snapshots
• Hadoop Flags: Reviewed
• Assignee: Haiping lv
• Reporter: Haiping lv
• Created: 31/Aug/23 15:17
• Updated: 19/Sep/23 06:31
• Resolved: 19/Sep/23 06:31

Description

HMaster encountered a deadlock when runing ITBLL test.

The handler thread of HMaster holds the synchronized lock of SnapshotManager and competes for the readLock of takingSnapshotLock.

Thread 57 (RpcServer.default.FPBQ.Fifo.handler=22,queue=1,port=16000):
  State: WAITING
  Blocked count: 14
  Waited count: 206
  Waiting on java.util.concurrent.locks.ReentrantReadWriteLock$FairSync@a7a58d4
  Stack:
    sun.misc.Unsafe.park(Native Method)
    java.util.concurrent.locks.LockSupport.park(LockSupport.java:175)
    java.util.concurrent.locks.AbstractQueuedSynchronizer.parkAndCheckInterrupt(AbstractQueuedSynchronizer.java:836)
    java.util.concurrent.locks.AbstractQueuedSynchronizer.doAcquireShared(AbstractQueuedSynchronizer.java:967)
    java.util.concurrent.locks.AbstractQueuedSynchronizer.acquireShared(AbstractQueuedSynchronizer.java:1283)
    java.util.concurrent.locks.ReentrantReadWriteLock$ReadLock.lock(ReentrantReadWriteLock.java:727)
    org.apache.hadoop.hbase.master.snapshot.SnapshotManager.takeSnapshot(SnapshotManager.java:675)
    org.apache.hadoop.hbase.master.MasterRpcServices.snapshot(MasterRpcServices.java:1765)
    org.apache.hadoop.hbase.shaded.protobuf.generated.MasterProtos$MasterService$2.callBlockingMethod(MasterProtos.java)
    org.apache.hadoop.hbase.ipc.RpcServer.call(RpcServer.java:437)
    org.apache.hadoop.hbase.ipc.CallRunner.run(CallRunner.java:124)
    org.apache.hadoop.hbase.ipc.RpcHandler.run(RpcHandler.java:102)
    org.apache.hadoop.hbase.ipc.RpcHandler.run(RpcHandler.java:82) 

The Master Snapshot cleaning thread holds the writeLock of takingSnapshotLock to compete for the synchronized lock of SnapshotManager.

Thread 5405 (hfile_cleaner-dir-scan-pool-2):
State: BLOCKED
Blocked count: 50535
Waited count: 92200
Blocked on org.apache.hadoop.hbase.master.snapshot.SnapshotManager@21a16019
Blocked by 57 (RpcServer.default.FPBQ.Fifo.handler=22,queue=1,port=16000)
Stack:
  org.apache.hadoop.hbase.master.snapshot.SnapshotManager.isTakingAnySnapshot(SnapshotManager.java:655)
  org.apache.hadoop.hbase.master.snapshot.SnapshotFileCache.getUnreferencedFiles(SnapshotFileCache.java:197)
  org.apache.hadoop.hbase.master.snapshot.SnapshotHFileCleaner.getDeletableFiles(SnapshotHFileCleaner.java:78)
  org.apache.hadoop.hbase.master.cleaner.CleanerChore.checkAndDeleteFiles(CleanerChore.java:375)
  org.apache.hadoop.hbase.master.cleaner.CleanerChore.lambda$traverseAndDelete$1(CleanerChore.java:483)
  org.apache.hadoop.hbase.master.cleaner.CleanerChore$$Lambda$788/1121007598.act(Unknown Source)
  org.apache.hadoop.hbase.master.cleaner.CleanerChore.deleteAction(CleanerChore.java:565)
  org.apache.hadoop.hbase.master.cleaner.CleanerChore.traverseAndDelete(CleanerChore.java:483)
  org.apache.hadoop.hbase.master.cleaner.CleanerChore.lambda$traverseAndDelete$2(CleanerChore.java:495)
  org.apache.hadoop.hbase.master.cleaner.CleanerChore$$Lambda$794/1250580839.run(Unknown Source)
  java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
  java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
  java.lang.Thread.run(Thread.java:750){{}} 

Issue Links

• is caused by: HBASE-26323 Introduce a SnapshotProcedure
• relates to: HBASE-27919 Run ITBLL for branch-3
• links to: GitHub Pull Request #5381