[
https://issues.apache.org/jira/browse/HBASE-10545?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
stack updated HBASE-10545:
--------------------------
Attachment: 10545.txt
This is an old issue provoked by my running tsdb. In the regionserver, after
all user regions are closed, if we are carrying the meta region, we will wait
until the request change rate is zero before we proceed to shut down. The idea
is we do not want to lose any writes that may be inbound around region close...
we want them all to land in meta before it goes down. With tsdb running, when
the user regions are gone, it keeps reading meta to try and find where all the
regions have gone.
Here is a patch that instead counts writes only. If no more writes, we go down
even though we may be taking reads on meta.
> RS Hangs waiting on region to close on shutdown; has to timeout before can go
> down
> ----------------------------------------------------------------------------------
>
> Key: HBASE-10545
> URL: https://issues.apache.org/jira/browse/HBASE-10545
> Project: HBase
> Issue Type: Bug
> Affects Versions: 0.98.0
> Reporter: stack
> Fix For: 0.96.2, 0.98.1, 0.99.0, 0.94.18
>
> Attachments: 10545.txt
>
>
> I am seeing cluster sometimes fails to go down hanging out waiting on close -
> it looks like it is waiting on hbase:meta that is the issue. I am running
> 0.98.0RC3 and hadoop-2.4.0-SNAPSHOT. Might be my setup. Filing this issue
> to keep an eye on this as I go.
> It looks like we are not calling close the region that is holding us up. Log
> is full of this:
> {code}
> 2014-02-14 16:07:21,095 DEBUG [regionserver60020] regionserver.HRegionServer:
> Waiting on 1588230740
> {code}
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
