[
https://issues.apache.org/jira/browse/HBASE-12229?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14167760#comment-14167760
]
Dima Spivak commented on HBASE-12229:
-------------------------------------
So createTable() in HBaseTestingUtility already has the
waitUntilAllRegionsAssigned. In this case, I just added the call to the wait
because HBTU didn't have a createTable method that takes splitKeys, but I can
add one of those instead of adding the call. More fundamentally, should we
consider whether HBA#createTable itself should be synchronous up through the
region assignments being made?
On a separate note, [~mbertozzi], is the
SnapshotTestingUtils#waitForTableToBeOnline right to be doing nothing with the
boolean returned by isTableAvailable? It looks like you might have created the
method and I wasn't sure whether to touch it with this JIRA.
> NullPointerException in SnapshotTestingUtils
> --------------------------------------------
>
> Key: HBASE-12229
> URL: https://issues.apache.org/jira/browse/HBASE-12229
> Project: HBase
> Issue Type: Bug
> Components: test
> Affects Versions: 1.0.0, 2.0.0, 0.98.7
> Reporter: Dima Spivak
> Assignee: Dima Spivak
> Priority: Minor
> Attachments: HBASE-12229.patch, HBASE-12229_v1.patch
>
>
> I tracked down occasional flakiness in TestRestoreSnapshotFromClient to a
> potential NPE in SnapshotTestingUtils#waitForTableToBeOnline. In short, some
> tests in TestRestoreSnapshot... create a table and then invoke
> SnapshotTestingUtils#waitForTableToBeOnline, but this method assumes that
> regions have been assigned by the time it's invoked (which is not always the
> case).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
