[
https://issues.apache.org/jira/browse/HBASE-9457?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13767881#comment-13767881
]
Francis Liu commented on HBASE-9457:
------------------------------------
Looked at the patch. It seems the zk namespace data gets stale with this
approach. Since the namespace table is lazily loaded, this causes the namespace
zk information to be lazily loaded as well. Which causes the data read by zkNM
clients to be stale until one of the namespace apis are invoked which seems to
be unbounded.
I think there are two other options:
1. move zkNM to a region observer coprocessor
2. roll namespace table into the meta handler classes (metaWAL, metaSSH, etc)
If time is a concern we can just remove zkNM and address it in a separate jira.
Good catch on the race condition with the janitor. I'll address that in
HBASE-9540.
> Master could fail start if region server with system table is down
> ------------------------------------------------------------------
>
> Key: HBASE-9457
> URL: https://issues.apache.org/jira/browse/HBASE-9457
> Project: HBase
> Issue Type: Bug
> Components: master, Region Assignment
> Reporter: Jimmy Xiang
> Assignee: Jimmy Xiang
> Priority: Critical
> Attachments: trunk-9457.patch, trunk-9457_v2.1.patch,
> trunk-9457_v2.2.patch, trunk-9457_v2.patch
>
>
> In the region server holding the system table is killed while master is
> starting, master will hang there waiting for system table to be assigned
> which won't happen.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
