[
https://issues.apache.org/jira/browse/HBASE-18415?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16099857#comment-16099857
]
Chia-Ping Tsai commented on HBASE-18415:
----------------------------------------
If the timeout occurs on the current request, we will get the
CellTimeoutException rather than the information (response or exception) from
server. In the other words, we don't know what happens in master. So there is a
problem for next "retry".
# Using different group/nonce may cause the Table***Exception if the previous
request succeeds and changes the state of targeted table.
# By contrast, using same group/nonce may receive the response with exception
if the previous request fails (the corresponding procId is recorded with
exception by master).
Will be back after more tests.
> The "retry" may cause TableExistsException even if the table doesn't exist
> --------------------------------------------------------------------------
>
> Key: HBASE-18415
> URL: https://issues.apache.org/jira/browse/HBASE-18415
> Project: HBase
> Issue Type: Bug
> Reporter: Chia-Ping Tsai
> Assignee: Chia-Ping Tsai
> Fix For: 3.0.0, 1.5.0, 2.0.0-alpha-2
>
> Attachments: HBASE-18415.branch-1.ut.patch,
> HBASE-18415.branch-1.v0.patch, HBASE-18415.branch-1.v1.patch
>
>
> After a timeout occurs on first request, client will retry the request with
> distinct group/nonce. The second request may bring the TableXXXException back
> if the first request have changed the table state.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
