[
https://issues.apache.org/jira/browse/HBASE-18357?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16095368#comment-16095368
]
Vladimir Rodionov edited comment on HBASE-18357 at 7/20/17 9:07 PM:
--------------------------------------------------------------------
Three tests fails:
h5. TestClusterStatus
The reason is described above: client connection is not getting periodic
ClusterStatus notifications. This functionality is totally missing in 2.0.
We can leave this test disabled or modify it to remove code which relied on
this functionality
h5. TestMulti
Gets deadlocked in Admin.move because test configures RPC_HANDLER_COUNT to be
equals to 1. One handler to Master is taken by test client and when RS tries to
report Master Region state change it is get blocked.
What we can do is to make all calls from RS to Master are HIGH priority calls
or to change test case and set RPC_HANDLER_COUNT to a default value (one test
in test suite relies on value of 1 though)
h5. TestRegionCaching
Fixed. Different semantics in 2.0 when call to a region fails due to stale info
in a region cache and there is only one attempt. In pre-2.0 HBase did not do
anything, in 2.0 HBase clears stale cache entry
[~syuanjiang], [~devaraj] what do you think?
was (Author: vrodionov):
Three tests fails:
h5. TestClusterStatus
The reason is described above: client connection is not getting periodic
ClusterStatus notifications. This functionality is totally missing in 2.0.
We can leave this test disabled or modify it to remove code which relied on
this functionality
h5. TestMulti
Gets deadlocked in Admin.move because test configures RPC_HANDLER_COUNT to be
equals to 1. One handler to Master is taken by test client and when RS tries to
report Master Region state change it is get blocked.
What we can do is to make all calls from RS to Master are HIGH priority calls
or to change test case and set RPC_HANDLER_COUNT to a default value (one test
in test suite relies on vale of 1 though)
h5. TestRegionCaching
Fixed. Different semantics in 2.0 when call to a region fails due to stale info
in a region cache and there is only one attempt. In pre-2.0 HBase did not do
anything, in 2.0 HBase clears stale cache entry
[~syuanjiang], [~devaraj] what do you think?
> Enable disabled tests in TestHCM that were disabled by Proc-V2 AM in
> HBASE-14614
> --------------------------------------------------------------------------------
>
> Key: HBASE-18357
> URL: https://issues.apache.org/jira/browse/HBASE-18357
> Project: HBase
> Issue Type: Bug
> Components: test
> Affects Versions: 2.0.0-alpha-1
> Reporter: Stephen Yuan Jiang
> Assignee: Vladimir Rodionov
>
> The Core Proc-V2 AM change in HBASE-14614 disabled two tests inTestHCM:
> testMulti and testRegionCaching
> This JIRA tracks the work to enable them.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
