hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ted Yu (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-8631) Meta Region First Recovery
Date Tue, 28 May 2013 20:43:20 GMT

    [ https://issues.apache.org/jira/browse/HBASE-8631?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13668649#comment-13668649
] 

Ted Yu commented on HBASE-8631:
-------------------------------

{code}
+      // we split meta regions and user regions separately therefore logfiles are either
all for
{code}
'meta regions' -> 'meta region'
{code}
+            if (carryingMeta && !region.equalsIgnoreCase(metaEncodeRegionName)) {
+              // skip non-meta regions because we recover meta regions and user regions in
+              // two separate calls.
+              continue;
+            } else if (!carryingMeta && region.equalsIgnoreCase(metaEncodeRegionName))
{
{code}
The conditions for above two checks can be unified, right ?
{code}
carryingMeta != region.equalsIgnoreCase(metaEncodeRegionName)
{code}
{code}
-        Set<ServerName> previouslyFailedDeletoins = null;
+        List<Pair<Set<ServerName>, Boolean>> previouslyFailedDeletoins
= null;
{code}
typo: Deletoin
{code}
+            // set a smaller retries to fast fail otherwise splitlogworker could be blocked
for
+            // quite a while inside HConnection layer. The worker won't available for other
+            // tasks even after current task is preempted after a split task times out.
{code}
Should the new conf be created inside LogReplayOutputSink ctor ?
                
> Meta Region First Recovery
> --------------------------
>
>                 Key: HBASE-8631
>                 URL: https://issues.apache.org/jira/browse/HBASE-8631
>             Project: HBase
>          Issue Type: Bug
>          Components: MTTR
>            Reporter: Jeffrey Zhong
>            Assignee: Jeffrey Zhong
>         Attachments: hbase-8631.patch
>
>
> We have a separate wal for meta region. While log splitting logic haven't taken the advantage
of this and splitlogworker still picks a wal file randomly. Imaging if we have multiple region
servers including meta RS fails about the same time while meta wal is recovered last, all
failed regions have to wait meta recovered and then can be online again. 
> The open JIRA is to let splitlogworker to pick a meta wal file firstly and then others.

--
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

Mime
View raw message