You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hbase.apache.org by "stack (JIRA)" <ji...@apache.org> on 2017/08/10 23:30:00 UTC
[jira] [Created] (HBASE-18562) [AMv2] expireServers and
ServerCrashProcedure cleanup
stack created HBASE-18562:
-----------------------------
Summary: [AMv2] expireServers and ServerCrashProcedure cleanup
Key: HBASE-18562
URL: https://issues.apache.org/jira/browse/HBASE-18562
Project: HBase
Issue Type: Bug
Components: Region Assignment
Reporter: stack
Priority: Critical
Fix For: 2.0.0
In review of HBASE-18551, [~uagashe] posed a scenario that revealed a hole in our processing of unassigns; there is case where a UP might not get notification from ServerCrashProcedure if the UP is scheduled AFTER a SCP has gotten past its handleRIT call (No new SCP will be queued because expireServer won't let it happen if crashed server is in dead server list which it will be).
Chatting on it, expireServers is doing checks that belong inside ServerCrashProcedure. expireServers scheduling an SCP each time it is called would make it so SCP processing is serialized one behind the other. If the first does the clean up all subsequent will do no work but Procedures dependent on them will get their wakeup call.
This issue is about implementing the above cleanup.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)