[
https://issues.apache.org/jira/browse/DERBY-4294?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12725981#action_12725981
]
Kristian Waagan commented on DERBY-4294:
----------------------------------------
I think this may be related to prefetching and the default heap size on
Windows. If the Sun JVM was used, I think the maximum heap size is set to 64 MB
(only client mode on Windows).
If I disable prefetching (derby.language.bulkFetchDefault=1) I'm able to run
the select with 10.3 using a 38 MB heap. With 10.6 I get away with a 37 MB heap.
I don't know the minimum overhead of Derby, but after running a values clause
in ij I see a heap usage of about 2 MB. After the select it rises to about 30
MB. Of these, around 26 MB are byte arrays, and I believe these are the page
cache. Since Derby is asked to produce strings from the byte arrays
representing the Clobs, we need heap space for these strings as well. If the
prefetch is set to 16 rows, all 14 rows will be fetched at once. 30 + 23 = 53
MB.
The remaining 11 MB must be data from the other columns and overhead. With
prefetch disabled, Derby/Java is able to discard some of the data during result
set processing.
It is not clear to me that there is anything wrong with the Clob handling in
this case.
> java.lang.OutOfMemoryError from mailjdbc test after running for one hour from
> select * from REFRESH.INBOX
> ---------------------------------------------------------------------------------------------------------
>
> Key: DERBY-4294
> URL: https://issues.apache.org/jira/browse/DERBY-4294
> Project: Derby
> Issue Type: Bug
> Components: Store
> Affects Versions: 10.3.3.0
> Environment: Window Vista
> Reporter: Lily Wei
> Attachments: Derby4294_pid6920.zip, Heapspaceerror.zip
>
> Original Estimate: 0h
> Remaining Estimate: 0h
>
> After running mailjdbc test on 10.3 branch, select * from REFRESH.INBOX gave
> me error like:
> ERROR XJ001: Java exception: 'Java heap space: java.lang.OutOfMemoryError'
> I also see the same error on 10.5 branch.
> I inculde my mailsdb directory in this bug too.
> How to reproduce this issue:
> 1. cd mailjdbc
> 2. java org.apache.derbyTesting.system.mailjdbc.MailJdbc embedded
> 3. Wait for an hour and ignore Primary key violation error
> ( : INFO :Refresh Thread : Error while inserting attachments:The statement
> was ab
> orted because it would have caused a duplicate key value in a unique or
> primary
> key constraint or unique index identified by 'ATTACH__PK' defined on
> 'ATTACH'.)
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
