[HBASE-15805] Canary test timeout if there is failed or shutdown zookeeper node in a quorum

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.98.20
• Fix Version/s: 0.98.20
• Component/s: canary
• Hadoop Flags: Reviewed
• Assignee: Vishal Khandelwal
• Reporter: Vishal Khandelwal
• Created: 09/May/16 11:32
• Updated: 05/Aug/16 11:24
• Resolved: 19/May/16 02:54

Description

Canary tools creates HBaseadmin object which which starts a thread to keep pinging zookeeper on a regular interval. This cause delay when one of the node in zookeeper is down and test timeouts.

Canary tool is creating HBaseadmin whose isEnabled and tableExists calls starts a catalog tacker --> retries for zookeeper connection. This zookeeper (6s) retry happens for multiple times for scanning 390+ region it is causing lot of sleeps. By removing the failed zookeeper from the test it actually took not more than 9 sec where it was taking almost 8 mins. Since "sniff" calls each table one by one thus causes the delay in the test.

2016-04-29 09:52:13 [Thread-6-SendThread()] DEBUG zookeeper.ClientCnxnSocketNIO(203): Ignoring exception during shutdown input
java.net.SocketException: Socket is not connected
at sun.nio.ch.Net.translateToSocketException(Net.java:123)
at sun.nio.ch.Net.translateException(Net.java:157)
at sun.nio.ch.Net.translateException(Net.java:163)
at sun.nio.ch.SocketAdaptor.shutdownInput(SocketAdaptor.java:401)
at org.apache.zookeeper.ClientCnxnSocketNIO.cleanup(ClientCnxnSocketNIO.java:200)
at org.apache.zookeeper.ClientCnxn$SendThread.cleanup(ClientCnxn.java:1246)
at org.apache.zookeeper.ClientCnxn$SendThread.cleanAndNotifyState(ClientCnxn.java:1236)
at org.apache.zookeeper.ClientCnxn$SendThread.run(ClientCnxn.java:1169)
Caused by: java.nio.channels.NotYetConnectedException
at sun.nio.ch.SocketChannelImpl.shutdownInput(SocketChannelImpl.java:782)
at sun.nio.ch.SocketAdaptor.shutdownInput(SocketAdaptor.java:399)
... 4 more
2016-04-29 09:52:13 [Thread-6-SendThread()] DEBUG zookeeper.ClientCnxnSocketNIO(210): Ignoring exception during shutdown output
java.net.SocketException: Socket is not connected
at sun.nio.ch.Net.translateToSocketException(Net.java:123)
at sun.nio.ch.Net.translateException(Net.java:157)
at sun.nio.ch.Net.translateException(Net.java:163)
at sun.nio.ch.SocketAdaptor.shutdownOutput(SocketAdaptor.java:409)
at org.apache.zookeeper.ClientCnxnSocketNIO.cleanup(ClientCnxnSocketNIO.java:207)
at org.apache.zookeeper.ClientCnxn$SendThread.cleanup(ClientCnxn.java:1246)
at org.apache.zookeeper.ClientCnxn$SendThread.cleanAndNotifyState(ClientCnxn.java:1236)
at org.apache.zookeeper.ClientCnxn$SendThread.run(ClientCnxn.java:1169)
Caused by: java.nio.channels.NotYetConnectedException
at sun.nio.ch.SocketChannelImpl.shutdownOutput(SocketChannelImpl.java:799)
at sun.nio.ch.SocketAdaptor.shutdownOutput(SocketAdaptor.java:407)