[ 
https://issues.apache.org/jira/browse/DERBY-5448?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13122182#comment-13122182
 ] 

Myrna van Lunteren commented on DERBY-5448:
-------------------------------------------

One more thing I thought to mention: I've not seen deadlocks in this test since 
I started running them (for QA cycle for 10.5, I think), but comments in the 
test lead me to believe they may have been possible.
I'm referring to this section in the Tester1 class:
                // Too many deadlocks amd locking issues if this is not 
commented out
                try {
                        connex.setAutoCommit(false);
                } catch (Exception e) {
                        System.out.println("FAIL: " + getThread_id()
                                        + "'s setAutoCommit() failed:");
                        printException("setting AutoCommit", e);
                }

                // also set isolation level to 
Connection.TRANSACTION_READ_UNCOMMITTED
                // to reduce number of
                // deadlocks
                setIsolationLevel(Connection.TRANSACTION_READ_UNCOMMITTED);
So perhaps a different hardware/jvm combination is more/less likely to run into 
deadlock situations...?

                
> In the 10.8 branch, back out the concurrency improvements introduced by using 
> SequenceUpdaters to allocate identity values.
> ---------------------------------------------------------------------------------------------------------------------------
>
>                 Key: DERBY-5448
>                 URL: https://issues.apache.org/jira/browse/DERBY-5448
>             Project: Derby
>          Issue Type: Bug
>          Components: Documentation, SQL
>    Affects Versions: 10.8.2.1
>            Reporter: Rick Hillegas
>             Fix For: 10.8.2.1
>
>         Attachments: derby-5448-01-aa-backoutConcurrencyChanges.diff, 
> derby-5448-02-aa-backoutDocsChanges.diff
>
>
> New errors were seen while running NsTest against the 10.8.2 release 
> candidates. These errors are recorded as DERBY-5430 and DERBY-5422. The first 
> error is a deadlock which occurs when inserting into the main table of the 
> test. The second error is a problem clearing the identity cache. These errors 
> have caused people to lose confidence in the concurrency improvements 
> introduced by DERBY-4437. We should back the DERBY-4437 changes out of the 
> 10.8 branch and use the trunk to continue debugging the problems disclosed by 
> NsTest.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: 
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Reply via email to