You are viewing a plain text version of this content. The canonical link for it is here.
Posted to derby-dev@db.apache.org by "Kristian Waagan (JIRA)" <ji...@apache.org> on 2010/04/18 16:14:25 UTC
[jira] Commented: (DERBY-4620) Query optimizer causes OOM error on
a complex query
[ https://issues.apache.org/jira/browse/DERBY-4620?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12858298#action_12858298 ]
Kristian Waagan commented on DERBY-4620:
----------------------------------------
I haven't had the time to investigate further, but on my laptop (Tecra M9, dual core Centrino) the repro runs okay:
2038 rows inserted/updated/deleted
ij>
real 0m21.677s
user 0m27.563s
sys 0m0.851s
$ java -version
java version "1.6.0_18"
Java(TM) SE Runtime Environment (build 1.6.0_18-b07)
Java HotSpot(TM) Server VM (build 16.0-b13, mixed mode)
Derby 10.5.3.0
With -Xmx128m and -client :
2038 rows inserted/updated/deleted
ij>
real 0m19.518s
user 0m17.211s
sys 0m0.498s
Can you run the repro on your side with -XX:+HeapDumpOnOutOfMemoryError and generate a heap histogram from the dump?
> Query optimizer causes OOM error on a complex query
> ---------------------------------------------------
>
> Key: DERBY-4620
> URL: https://issues.apache.org/jira/browse/DERBY-4620
> Project: Derby
> Issue Type: Bug
> Components: SQL
> Affects Versions: 10.5.3.0
> Environment: java version "1.6.0_17"
> Java(TM) SE Runtime Environment (build 1.6.0_17-b04)
> Java HotSpot(TM) Client VM (build 14.3-b01, mixed mode, sharing)
> Linux rocio.xxx 2.6.24-27-generic #1 SMP Fri Mar 12 01:10:31 UTC 2010 i686 GNU/Linux
> Reporter: Chris Wilson
>
> I have a query that generates about 2,000 rows in a summary table. It runs fast enough (~30 seconds) on MySQL. The same query on Derby runs for about 10 minutes and then fails with an OutOfMemoryError.
> I have created a test case to reproduce the problem. It's not minimal because it relies on a rather large dataset to reproduce it, and it's not trivial, but I don't mind doing a bit of work trimming it if someone can point me in the necessary direction.
> You can check out the test case from our Subversion server here:
> http://rita.wfplogistics.org/svn/trackrita/rita/doc/derby-oom-slow-query
> which includes a pre-built Derby database in "testdb.derby". If this database is deleted, "test.sh" will recreate it, but that takes about 10-15 minutes.
> Just modify the script "test.sh" to point to your Derby libraries, and run it (or just execute the commands in "movement_complete.sql") to demonstrate the problem. You can view the source of that file online here:
> http://rita.wfplogistics.org/trac/browser/rita/conf/movement_complete.sql
> The first "insert into movement_complete" (starting around line 4) takes about 15 seconds to complete and inserts 5890 rows. The second, starting around line 54, does not complete in reasonable time on Derby. On MySQL, it runs in 28 seconds and inserts 2038 rows. On Derby, after 10 minutes I get:
> JAVA ERROR: java.lang.OutOfMemoryError: Java heap space
> ij> ERROR X0Y67: Cannot issue rollback in a nested connection when there
> is a pending operation in the parent connection.
> (process exits)
> It does not output the query plan in this case.
> Following the suggestion of Bryan Pendleton, I tried increasing the JVM memory limit from the default to 1024m, and this allows the query to finish executing quite quickly. I guess that means that the optimiser is just taking a lot of memory to
> optimise the query, and it spends forever in GC before finally hitting OOM and giving up when using the default settings.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: https://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira