You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "Mike Drob (JIRA)" <ji...@apache.org> on 2017/09/27 19:48:00 UTC

[jira] [Commented] (HBASE-18247) Hbck to fix the case that replica region shows as key in the meta table

    [ https://issues.apache.org/jira/browse/HBASE-18247?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16183141#comment-16183141 ] 

Mike Drob commented on HBASE-18247:
-----------------------------------

Is this relevant for master/branch-2? Those are getting a huge hbck rewrite, aren't they?

> Hbck to fix the case that replica region shows as key in the meta table
> -----------------------------------------------------------------------
>
>                 Key: HBASE-18247
>                 URL: https://issues.apache.org/jira/browse/HBASE-18247
>             Project: HBase
>          Issue Type: Bug
>          Components: hbck, Operability
>    Affects Versions: 2.0.0-alpha-1
>            Reporter: huaxiang sun
>            Assignee: huaxiang sun
>            Priority: Minor
>             Fix For: 1.4.0, 1.3.2, 1.5.0, 1.2.7
>
>         Attachments: HBASE-18247-branch-1.2-v001.patch
>
>
> Recently, we run into one case with read replica, the replica region shows up as key in meta table (it is not supposed to happen, we are still working on why it showed up in the meta table).
> However, hbck always reported the error about the primary region. Please see the error attached.
> {code}
> The entry in meta table
> test,92b0201b,1492546349354_0001.c3e6f235fe7caef75f8b0fb92a012da3. column=info:regioninfo, timestamp=1494958820573, value={ENCODED => c3e6f235fe7caef75f8b0fb92a012da3, NAME => 'test,92b0201b,1492546349354_0001.c3e6f235fe7caef75f8b0fb92a012da3.', STARTKEY => '92b0201b', ENDKEY => '92f1a952', REPLICA_ID => 1}
> ERROR: Region { meta => test,92b0201b,1492546349354.d2c637715f31a072f174e70d407fb458., hdfs => null, deployed => , replicaId => 0 } found in META, but not in HDFS or deployed on any region server.
> {code}
> Traced the code, in the following line, it does not consider the case that replicaId in regionInfo could be non-default. 
> https://github.com/apache/hbase/blob/master/hbase-client/src/main/java/org/apache/hadoop/hbase/MetaTableAccessor.java#L985
> If it is changed to get replicaId from regionInfo, then hbck should be able to fix this by "-fixMeta".
> {code}
> diff --git a/hbase-client/src/main/java/org/apache/hadoop/hbase/MetaTableAccessor.java b/hbase-client/src/main/java/org/apache/hadoop/hbase/MetaTableAccessor.java
> index 9eb5111..1649e53 100644
> --- a/hbase-client/src/main/java/org/apache/hadoop/hbase/MetaTableAccessor.java
> +++ b/hbase-client/src/main/java/org/apache/hadoop/hbase/MetaTableAccessor.java
> @@ -982,7 +982,7 @@ public class MetaTableAccessor {
>      List<HRegionLocation> locations = new ArrayList<>(1);
>      NavigableMap<byte[],NavigableMap<byte[],byte[]>> familyMap = r.getNoVersionMap();
>  
> -    locations.add(getRegionLocation(r, regionInfo, 0));
> +    locations.add(getRegionLocation(r, regionInfo, regionInfo.getReplicaId()));
>  
>      NavigableMap<byte[], byte[]> infoMap = familyMap.get(getCatalogFamily());
>      if (infoMap == null) return new RegionLocations(locations);
> {code}



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)