[
https://issues.apache.org/jira/browse/HBASE-3007?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12910341#action_12910341
]
Nicolas Spiegelberg commented on HBASE-3007:
--------------------------------------------
All instances of over-sizing happen during compactions. One particularly
egregious one...
{code}
2010-09-16 11:38:51,066 INFO org.apache.hadoop.hbase.regionserver.Store:
Started compaction of 4 file(s) in Actions of
dev,55555550,1284591819438.6774c3a858b607148c92469d0c850bef. into
hdfs://snc4.facebook.com:9000/SNC4-HBASE/ev/6774c3a858b607148c92469d0c850bef/.tmp,
seqid=1837996
2010-09-16 11:38:51,242 INFO org.apache.hadoop.hbase.regionserver.StoreFile:
Bloom added to HFile. 256B, 3020/156 (1,936%)
{code}
> StoreFile Blooms are Being Stored Undersized
> --------------------------------------------
>
> Key: HBASE-3007
> URL: https://issues.apache.org/jira/browse/HBASE-3007
> Project: HBase
> Issue Type: Bug
> Components: regionserver
> Affects Versions: 0.89.20100621
> Reporter: Nicolas Spiegelberg
> Assignee: Nicolas Spiegelberg
> Priority: Minor
>
> While looking through error logs today, I noticed the following line.
> {code}
> 2010-09-16 04:12:26,401 INFO org.apache.hadoop.hbase.regionserver.StoreFile:
> Bloom added to HFile. 9600B, 10292/6933 (148%)
> {code}
> The last 3 numbers are: # of keys in bloom, # of keys preallocated for bloom,
> % full. The last number should never be > 100%. Oversized blooms will cause
> increased false positives. Note that this occurred on a Row+Col bloom. Will
> provide more details after further instrumentation.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
