hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Francis Liu (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-11546) Backport HBASE-11059 to 0.98
Date Fri, 18 Jul 2014 23:02:15 GMT

    [ https://issues.apache.org/jira/browse/HBASE-11546?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14067053#comment-14067053

Francis Liu commented on HBASE-11546:

Let's take a step back. Why is this change needed in 0.98?
Because 0.98 is the latest stable branch. If this enhancement proves itself useful then we
will be running this in production within the next few months.

Also, on any minor upgrade e.g. 0.98.4 to 0.98.5, it should be possible to do a rolling restart
from .4 to .5. This change would seems to require two rolling restarts, and on the second
restart the ordering of the master restart will be important.
Since this is an opt-in feature then requiring 2 rolling restarts should be okay?

Without a backport of HBASE-10569, which I don't think we can backport to a stable branch,
meta updates after this patch is applied in the current state would happen via a remote RPC
call in a synchronized block as per my understanding, which won't be ok.
Off-hand I don't see why colocation would be required. We'll take a look if we can come up
with a less invasive solution to avoid the heavy synchronization.

> Backport HBASE-11059 to 0.98
> ----------------------------
>                 Key: HBASE-11546
>                 URL: https://issues.apache.org/jira/browse/HBASE-11546
>             Project: HBase
>          Issue Type: Task
>            Reporter: Virag Kothari
> Discuss concerns about backporting HBASE-11059 to 0.98

This message was sent by Atlassian JIRA

View raw message