[
https://issues.apache.org/jira/browse/HBASE-15591?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15227730#comment-15227730
]
Matteo Bertozzi commented on HBASE-15591:
-----------------------------------------
a yield is supposed to let someone else run. so adding a sleep on yield is not
a valid generic solution.
The idea is that in similar cases where you have to wait on something you go in
a waiting queue until you get the event.
but we still not have everything to implement that, so at the moment this
procedure does not follow the "procedure best practice".
and it creates bad behaviors.
you can add the sleep before throwing the yield exception if you want slow down
the retries
> ServerCrashProcedure not yielding
> ---------------------------------
>
> Key: HBASE-15591
> URL: https://issues.apache.org/jira/browse/HBASE-15591
> Project: HBase
> Issue Type: Bug
> Affects Versions: 2.0.0, 1.2.0
> Reporter: Jerry He
> Assignee: Jerry He
> Attachments: HBASE-15591.patch
>
>
> ServerCrashProcedure is not propagating ProcedureYieldException to the
> ProcedureExecutor
> One symptom is that while ServerCrashProcedure is waiting for meta to be up
> the Procedure WALs get filled up rapidly with all the executions.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
