[
https://issues.apache.org/jira/browse/HBASE-2922?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12900903#action_12900903
]
HBase Review Board commented on HBASE-2922:
-------------------------------------------
Message from: "Jean-Daniel Cryans" <[email protected]>
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://review.cloudera.org/r/694/
-----------------------------------------------------------
(Updated 2010-08-20 15:51:17.366257)
Review request for hbase.
Changes
-------
New patch, I decided to dig even more into this problem as I'm thinking that
what happens in the updateLock should be as quick as possible. The difference
with the previous patch is that the new HLog is created before we take the
updateLock, as well as all the preparation surrounding that. Then I tested on a
randomWrite 4 + 6 incrementing threads on a single RS that was instrumented to
sysout the time it took to go through the sync block in rollWriter.
Without the patch:
Test duration: 406835ms
Total sync block duration: 56048ms
Percentage of the time updates were blocked: 13.7%
Average time in sync block: 610ms
Median time in sync block: 37ms
With the patch:
Test duration: 383503ms
Total sync block duration: 14072ms
Percentage of the time updates were blocked: 3.6%
Average time in sync block: 142ms
Median time in sync block: 3ms
Comments:
I was surprised by how closing a HLog could take a wildly varying time. In both
cases I saw multi-seconds to close a writer (as you can tell by the averages
compared to medians). In any case, my patch makes it much faster and less
blocking. Your mileage may vary.
Summary
-------
Simply moves the LogActionListeners and outputfiles handling out of the
updateLock synchronization.
This addresses bug HBASE-2922.
http://issues.apache.org/jira/browse/HBASE-2922
Diffs (updated)
-----
/trunk/src/main/java/org/apache/hadoop/hbase/regionserver/wal/HLog.java
987355
Diff: http://review.cloudera.org/r/694/diff
Testing
-------
Unit tests and some PEs.
Thanks,
Jean-Daniel
> HLog cleanup is done under the updateLock, major slowdown
> ---------------------------------------------------------
>
> Key: HBASE-2922
> URL: https://issues.apache.org/jira/browse/HBASE-2922
> Project: HBase
> Issue Type: Bug
> Affects Versions: 0.20.6, 0.89.20100621
> Reporter: Jean-Daniel Cryans
>
> Something I've seen quite often in our production environment:
> {quote}
> 2010-08-16 16:17:27,104 INFO org.apache.hadoop.hbase.regionserver.HLog:
> removing old hlog file
> /hbase/.logs/rs22,60020,1280909840873/hlog.dat.1282000385321 whose highest
> sequence/edit id is 64837079950
> 2010-08-16 16:17:27,286 INFO org.apache.hadoop.hbase.regionserver.HLog:
> removing old hlog file
> /hbase/.logs/rs22,60020,1280909840873/hlog.dat.1282000392770 whose highest
> sequence/edit id is 64837088260
> 2010-08-16 16:17:27,452 INFO org.apache.hadoop.hbase.regionserver.HLog:
> removing old hlog file
> /hbase/.logs/rs22,60020,1280909840873/hlog.dat.1282000399300 whose highest
> sequence/edit id is 64837096566
> 2010-08-16 16:17:27,635 INFO org.apache.hadoop.hbase.regionserver.HLog:
> removing old hlog file
> /hbase/.logs/rs22,60020,1280909840873/hlog.dat.1282000406997 whose highest
> sequence/edit id is 64837104865
> 2010-08-16 16:17:27,827 INFO org.apache.hadoop.hbase.regionserver.HLog:
> removing old hlog file
> /hbase/.logs/rs22,60020,1280909840873/hlog.dat.1282000413803 whose highest
> sequence/edit id is 64837113153
> 2010-08-16 16:17:27,993 INFO org.apache.hadoop.hbase.regionserver.HLog:
> removing old hlog file
> /hbase/.logs/rs22,60020,1280909840873/hlog.dat.1282000421709 whose highest
> sequence/edit id is 64837121467
> 2010-08-16 16:17:28,160 INFO org.apache.hadoop.hbase.regionserver.HLog:
> removing old hlog file
> /hbase/.logs/rs22,60020,1280909840873/hlog.dat.1282000427333 whose highest
> sequence/edit id is 64837129775
> 2010-08-16 16:17:28,432 INFO org.apache.hadoop.hbase.regionserver.HLog:
> removing old hlog file
> /hbase/.logs/rs22,60020,1280909840873/hlog.dat.1282000434365 whose highest
> sequence/edit id is 64837138074
> 2010-08-16 16:17:28,518 INFO org.apache.hadoop.hbase.regionserver.HLog:
> removing old hlog file
> /hbase/.logs/rs22,60020,1280909840873/hlog.dat.1282000440347 whose highest
> sequence/edit id is 64837146376
> 2010-08-16 16:17:28,612 WARN org.apache.hadoop.hbase.regionserver.HLog: IPC
> Server handler 39 on 60020 took 1801ms appending an edit to hlog; editcount=0
> 2010-08-16 16:17:28,615 WARN org.apache.hadoop.hbase.regionserver.HLog: IPC
> Server handler 37 on 60020 took 1804ms appending an edit to hlog; editcount=1
> 2010-08-16 16:17:28,615 WARN org.apache.hadoop.hbase.regionserver.HLog: IPC
> Server handler 25 on 60020 took 1805ms appending an edit to hlog; editcount=2
> ...
> 2010-08-16 16:17:28,619 WARN org.apache.hadoop.hbase.regionserver.HLog: IPC
> Server handler 41 on 60020 took 1875ms appending an edit to hlog; editcount=50
> 2010-08-16 16:17:28,619 WARN org.apache.hadoop.hbase.regionserver.HLog: IPC
> Server handler 24 on 60020 took 1876ms appending an edit to hlog; editcount=51
> 2010-08-16 16:17:28,619 WARN org.apache.hadoop.hbase.regionserver.HLog: IPC
> Server handler 48 on 60020 took 1881ms appending an edit to hlog; editcount=54
> {quote}
> And looking at HLog.rollWriter, we roll then cleanup those unused hlog files
> under updateLock, which blocks all the appenders (as shown). We should only
> do the first part under that lock
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
