[
https://issues.apache.org/jira/browse/HBASE-12162?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14157612#comment-14157612
]
Ted Yu commented on HBASE-12162:
--------------------------------
The new test would establish MasterKeepAliveConnection and before
getHTableDescriptor() is called on this connection, kill the corresponding
master.
Going over existing tests, such as TestAdmin, I didn't find similar test case.
How about adding new test which verifies the robustness of all calls involving
MasterCallable in a separate issue ?
> HBaseAdmin#getTableDescriptor() may fail in case master fails over
> ------------------------------------------------------------------
>
> Key: HBASE-12162
> URL: https://issues.apache.org/jira/browse/HBASE-12162
> Project: HBase
> Issue Type: Bug
> Reporter: Ted Yu
> Assignee: Ted Yu
> Attachments: 12162-v1.txt
>
>
> This was discovered by Chakradhar Medavarapu during HA testing.
> Here is relevant exception:
> {code}
> 2014-09-30 04:07:56,734|beaver.machine|INFO|5728|5604|MainThread|14/09/30
> 04:07:56 ERROR util.AbstractHBaseTool: Error running command-line tool
> 2014-09-30
> 04:07:56,734|beaver.machine|INFO|5728|5604|MainThread|java.io.IOException:
> Call to onprem-ha34/10.215.18.85:60000 failed on local exception:
> java.io.IOException: Call id=1, waitTime=8703
> 2014-09-30 04:07:56,734|beaver.machine|INFO|5728|5604|MainThread|at
> org.apache.hadoop.hbase.ipc.RpcClient.wrapException(RpcClient.java:1571)
> 2014-09-30 04:07:56,734|beaver.machine|INFO|5728|5604|MainThread|at
> org.apache.hadoop.hbase.ipc.RpcClient.call(RpcClient.java:1541)
> 2014-09-30 04:07:56,736|beaver.machine|INFO|5728|5604|MainThread|at
> org.apache.hadoop.hbase.ipc.RpcClient.callBlockingMethod(RpcClient.java:1723)
> 2014-09-30 04:07:56,736|beaver.machine|INFO|5728|5604|MainThread|at
> org.apache.hadoop.hbase.ipc.RpcClient$BlockingRpcChannelImplementation.callBlockingMethod(RpcClient.java:1776)
> 2014-09-30 04:07:56,736|beaver.machine|INFO|5728|5604|MainThread|at
> org.apache.hadoop.hbase.protobuf.generated.MasterProtos$MasterService$BlockingStub.getTableDescriptors(MasterProtos.java:42525)
> 2014-09-30 04:07:56,736|beaver.machine|INFO|5728|5604|MainThread|at
> org.apache.hadoop.hbase.client.ConnectionManager$HConnectionImplementation$5.getTableDescriptors(ConnectionManager.java:2121)
> 2014-09-30 04:07:56,736|beaver.machine|INFO|5728|5604|MainThread|at
> org.apache.hadoop.hbase.client.ConnectionManager$HConnectionImplementation.getHTableDescriptor(ConnectionManager.java:2600)
> 2014-09-30 04:07:56,736|beaver.machine|INFO|5728|5604|MainThread|at
> org.apache.hadoop.hbase.client.HBaseAdmin.getTableDescriptor(HBaseAdmin.java:410)
> {code}
> From stack trace, exception came out of connection.getHTableDescriptor().
> This happened during master failover where MasterKeepAliveConnection to the
> failed master became unusable.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
