hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Pavel Salimov (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HBASE-18098) Region data ended up in a file of an another region
Date Wed, 24 May 2017 11:22:04 GMT

     [ https://issues.apache.org/jira/browse/HBASE-18098?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Pavel Salimov updated HBASE-18098:
----------------------------------
    Description: 
Somehow it happened that a region, say AB, was split onto A and B, but some portion of B's
data were missing in B's H-files, still presenting in A's files. I am not completely sure,
was it a result of just a failure during split, or of applying hbck repair after that. 

Anyway, data of some rows belonging to B were missing accessing normally. I was able to access
them making a scan type query regarding these rows to region A (despite the rows are after
its end). Hbck repairs, splitting A onto A0, A1 and (full) compaction did not change the situation:
the data were still missing in B migrated to A1. 

Copying the file from A1 dir to B finally made the data accessible.
 
I am reporting the issue in hope to get an advice on how to detect such an inconsistency as
well as hoping to clarify what leaded to such a state (and fixed if it is a bug).

  was:
Somehow it happened that a region, say AB, was split onto A and B, but some portion of B's
data were missing in B's H-files, still presenting in A's files. I am not completely sure,
was it a result of just a failure during split, or of applying hbck repair after that. 

Anyway, data of some rows belonging to B were missing accessing normally. I was able to access
them making a scan type query regarding these rows to region A (despite the rows are after
its end). Hbck repairs, splitting A onto A0, A1 and (full) compaction did not change the situation:
the data were still missing in B migrated to A1. 
 
I am reporting the issue in hope to get an advice on how to detect such an inconsistency as
well as hoping to clarify what leaded to such a state (and fixed if it is a bug).


> Region data ended up in a file of an another region
> ---------------------------------------------------
>
>                 Key: HBASE-18098
>                 URL: https://issues.apache.org/jira/browse/HBASE-18098
>             Project: HBase
>          Issue Type: Bug
>    Affects Versions: 1.2.0
>            Reporter: Pavel Salimov
>
> Somehow it happened that a region, say AB, was split onto A and B, but some portion of
B's data were missing in B's H-files, still presenting in A's files. I am not completely sure,
was it a result of just a failure during split, or of applying hbck repair after that. 
> Anyway, data of some rows belonging to B were missing accessing normally. I was able
to access them making a scan type query regarding these rows to region A (despite the rows
are after its end). Hbck repairs, splitting A onto A0, A1 and (full) compaction did not change
the situation: the data were still missing in B migrated to A1. 
> Copying the file from A1 dir to B finally made the data accessible.
>  
> I am reporting the issue in hope to get an advice on how to detect such an inconsistency
as well as hoping to clarify what leaded to such a state (and fixed if it is a bug).



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

Mime
View raw message