[
https://issues.apache.org/jira/browse/HBASE-18254?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16058714#comment-16058714
]
stack commented on HBASE-18254:
-------------------------------
I was worried that this wait on hbase:meta to be loaded would make a case where
if the crashed server was carrying hbase:meta, then we'd never start the SCP
(because it wouldn't start until meta was up... but its supposed to put up
hbase:meta). As it happens, HMaster will recover hbase:meta special outside of
SCP before it will let SCPs run. See this section in Master:
// get a list for previously failed RS which need log splitting work
// we recover hbase:meta region servers inside master initialization and
// handle other failed servers in SSH in order to start up master node ASAP
MasterMetaBootstrap metaBootstrap = createMetaBootstrap(this, status);
metaBootstrap.splitMetaLogsBeforeAssignment();
... So, this patch is good IMO. Let me commit.
> ServerCrashProcedure checks and waits for meta initialized, instead should
> check and wait for meta loaded
> ---------------------------------------------------------------------------------------------------------
>
> Key: HBASE-18254
> URL: https://issues.apache.org/jira/browse/HBASE-18254
> Project: HBase
> Issue Type: Bug
> Components: amv2
> Affects Versions: 2.0.0-alpha-1
> Reporter: Umesh Agashe
> Assignee: Umesh Agashe
> Fix For: 2.0.0-alpha-2
>
> Attachments: HBASE-18254.master.001.patch
>
>
> After enabling test
> hbase.master.procedure.TestServerCrashProcedure#testRecoveryAndDoubleExecutionOnRsWithMeta,
> this bug is found in ServerCrashProcedure.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
