[jira] Commented: (HBASE-1784) Missing rows after medium intensity insert

["stack (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HBASE-1784?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12749673#action_12749673 ] 

stack commented on HBASE-1784:
------------------------------

>From the log posted by Mathias, this is the pertinent bit:
{code}
2009-08-31 17:03:11,689 INFO org.apache.hadoop.hbase.master.RegionServerOperation: domirama,00AZRPXURYW7\x13\x7B_\x7F\xFF\xFF\xFE,1251730985136 open on 10.154.99.183:60020
2009-08-31 17:03:11,690 INFO org.apache.hadoop.hbase.master.RegionServerOperation: Updated row domirama,00AZRPXURYW7\x13\x7B_\x7F\xFF\xFF\xFE,1251730985136 in region .META.,,1 with startcode=1251704407620, server=10.154.99.183:60020
2009-08-31 17:03:11,749 INFO org.apache.hadoop.hbase.master.ServerManager: 5 region servers, 0 dead, average load 50.0
@@2009-08-31 17:03:11,851 DEBUG org.apache.hadoop.hbase.master.BaseScanner: Current assignment of domirama,00AZRPXUPUF7\xFC\xD4\xDF\x7F\xFF\xFF\xFC,1251730985136 is not valid;  serverAddress=, startCode=0 unknown.
{code}

Here, the split has been opened successfully out on the regionserver, so its state in the master has been cleared... but BaseScanner which has been running during the update sees old state of this row, not the state had been updated 150ms earlier.

Let me go through all places where we set a region unassigned and add in a get so we get current state rather than a stale state of a row.  I did this for the special case of splits in the previous patch applied against this issue but looks like it needs to be done more generally.


> Missing rows after medium intensity insert
> ------------------------------------------
>
>                 Key: HBASE-1784
>                 URL: https://issues.apache.org/jira/browse/HBASE-1784
>             Project: Hadoop HBase
>          Issue Type: Bug
>    Affects Versions: 0.20.0
>            Reporter: Jean-Daniel Cryans
>            Priority: Blocker
>             Fix For: 0.20.0
>
>         Attachments: 1784.patch, DataLoad.java, dbl-assignment-20090831, double-assignment, HBASE-1784-StoreFileScanner-hack.patch, HBASE-1784.log, META.log, processSplitRegion-check-regionIsOpening.patch
>
>
> This bug was uncovered by Mathias in his mail "Issue on data load with 0.20.0-rc2". Basically, somehow, after a medium intensity insert a lot of rows goes missing. Easy way to reproduce : PE. Doing a PE scan or randomRead afterwards won't uncover anything since it doesn't bother about null rows. Simply do a count in the shell, easy to test (I changed my scanner caching in the shell to do it faster).
> I tested some light insertions with force flush/compact/split in the shell and it doesn't break.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.