Alex Rudyy created QPID-8366:
--------------------------------
Summary: [Broker-J] The loss of BDB HA majority on invocation of
house keeping operations can crash the broker
Key: QPID-8366
URL: https://issues.apache.org/jira/browse/QPID-8366
Project: Qpid
Issue Type: Task
Components: Broker-J
Affects Versions: qpid-java-broker-7.1.4, qpid-java-broker-7.1.3,
qpid-java-broker-7.0.8, qpid-java-broker-7.1.2, qpid-java-broker-7.1.1,
qpid-java-broker-7.0.7, qpid-java-broker-7.0.6, qpid-java-broker-7.0.5,
qpid-java-broker-7.0.4, qpid-java-broker-7.1.0
Reporter: Alex Rudyy
The {{ConnectionScopedRuntimeException}} thrown from {{VirtualHost}} {{House
Keeping}} thread on invocation of {{MessageStore}} operations like
{{checkMessageStatus}} can crash the broker. An example of such exception stack
trace (from Qpid Broker version 7.0.6) is provided below:
{noformat}
2019-09-27 07:53:38,168 ERROR [virtualhost-test-pool-1] (o.a.q.s.Main) -
Uncaught exception, shutting down.
org.apache.qpid.server.util.ConnectionScopedRuntimeException: Required number
of nodes not reachable
at
org.apache.qpid.server.store.berkeleydb.replication.ReplicatedEnvironmentFacade.handleDatabaseException(ReplicatedEnvironmentFacade.java:495)
at
org.apache.qpid.server.store.berkeleydb.replication.ReplicatedEnvironmentFacade.commit(ReplicatedEnvironmentFacade.java:332)
at
org.apache.qpid.server.store.berkeleydb.AbstractBDBMessageStore.removeMessage(AbstractBDBMessageStore.java:288)
at
org.apache.qpid.server.store.berkeleydb.AbstractBDBMessageStore$StoredBDBMessage.remove(AbstractBDBMessageStore.java:1090)
at
org.apache.qpid.server.message.AbstractServerMessageImpl.decrementReference(AbstractServerMessageImpl.java:118)
at
org.apache.qpid.server.message.AbstractServerMessageImpl.access$500(AbstractServerMessageImpl.java:37)
at
org.apache.qpid.server.message.AbstractServerMessageImpl$Reference.release(AbstractServerMessageImpl.java:309)
at
org.apache.qpid.server.queue.QueueEntryImpl.dispose(QueueEntryImpl.java:557)
at
org.apache.qpid.server.queue.QueueEntryImpl.delete(QueueEntryImpl.java:572)
at
org.apache.qpid.server.queue.AbstractQueue$11.postCommit(AbstractQueue.java:1729)
at
org.apache.qpid.server.txn.AutoCommitTransaction.dequeue(AutoCommitTransaction.java:92)
at
org.apache.qpid.server.queue.AbstractQueue.dequeueEntry(AbstractQueue.java:1722)
at
org.apache.qpid.server.queue.AbstractQueue.dequeueEntry(AbstractQueue.java:1717)
at
org.apache.qpid.server.queue.AbstractQueue.deleteEntry(AbstractQueue.java:1761)
at
org.apache.qpid.server.queue.AbstractQueue.checkMessageStatus(AbstractQueue.java:2165)
at
org.apache.qpid.server.virtualhost.AbstractVirtualHost$VirtualHostHouseKeepingTask.execute(AbstractVirtualHost.java:1965)
at
org.apache.qpid.server.virtualhost.HouseKeepingTask$1.run(HouseKeepingTask.java:56)
at java.security.AccessController.doPrivileged(Native Method)
at
org.apache.qpid.server.virtualhost.HouseKeepingTask.run(HouseKeepingTask.java:51)
at
java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
at java.util.concurrent.FutureTask.runAndReset(FutureTask.java:308)
at
java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.access$301(ScheduledThreadPoolExecutor.java:180)
at
java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.run(ScheduledThreadPoolExecutor.java:294)
at
java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at
java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at
org.apache.qpid.server.bytebuffer.QpidByteBufferFactory.lambda$null$0(QpidByteBufferFactory.java:464)
at java.lang.Thread.run(Thread.java:748)
Caused by: com.sleepycat.je.rep.InsufficientAcksException: (JE 7.4.5)
Transaction: -3459038252 VLSN: 10,380,435,448, initiated at: 07:53:20.
Insufficient acks for policy:SIMPLE_MAJORITY. Need replica acks: 2. Missing
replica acks: 2. Timeout: 15000ms. FeederState=acc3_2(3)[MASTER]
Current feeds:
acc3_1: feederVLSN=10,380,435,456 replicaTxnEndVLSN=10,380,435,396
acc3: feederVLSN=10,380,435,456 replicaTxnEndVLSN=10,380,435,396
at
com.sleepycat.je.rep.impl.node.DurabilityQuorum.ensureSufficientAcks(DurabilityQuorum.java:205)
at
com.sleepycat.je.rep.stream.FeederTxns.awaitReplicaAcks(FeederTxns.java:189)
at
com.sleepycat.je.rep.impl.RepImpl.postLogCommitHookInternal(RepImpl.java:1426)
at
com.sleepycat.je.rep.impl.RepImpl.postLogCommitHook(RepImpl.java:1385)
at
com.sleepycat.je.rep.txn.MasterTxn.postLogCommitHook(MasterTxn.java:228)
at com.sleepycat.je.txn.Txn.commit(Txn.java:772)
at com.sleepycat.je.Transaction.doCommit(Transaction.java:621)
at com.sleepycat.je.Transaction.commit(Transaction.java:401)
at
org.apache.qpid.server.store.berkeleydb.replication.ReplicatedEnvironmentFacade.commit(ReplicatedEnvironmentFacade.java:328)
... 25 common frames omitted
{noformat}
The issue reported with the stack trace above occurred when BDB HA
{{VirtualHost}} was trying to delete an expired message, but its BDB HA group
lost the majority when the {{VirtualHost}} tried to commit a BDB HA
transaction for message deletion operation. The majority loss is communicated
as {{ConnectionScopeRuntimeException}} to the caller. It seems we need to catch
and handle {{ConnectionScopeRuntimeException}} in House Keeping operations.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
