db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Kristian Waagan <kristian.waa...@oracle.com>
Subject Re: istat-related problems seen in the regression tests
Date Fri, 18 Feb 2011 22:39:27 GMT
On 18.02.2011 20:31, Rick Hillegas wrote:
> On 2/18/11 11:16 AM, Lily Wei wrote:
>> Hi Rick:
>> On Fri, Feb 18, 2011 at 7:17 AM, Rick Hillegas 
>> <rick.hillegas@oracle.com <mailto:rick.hillegas@oracle.com>> wrote:
>>
>>     I have logged DERBY-5057 to track an out-of-memory error seen in
>>     the istat tests on 1.5 during the nightly run last night:
>>     
>> http://dbtg.foundry.sun.com/derby/test/Daily/jvm1.5/testing/Limited/testSummary-1071310.html
>>
>>     The current roster of known istat-related bugs is:
>>
>>     o DERBY-4940: A placeholder for discussion about and improvements
>>     to the default istat settings.
>>
>>     o DERBY-5030: May have been fixed by the patch committed on
>>     DERBY-5037. I didn't see this error in last night's nightly test
>>     run. Is anyone still seeing this bug? Lily, you mentioned that you
>>     had seen it after the DERBY-5037 patch was committed. Are you
>>     still seeing it?
>>
>>  Yes, it is a intermittent failure and I am still seeing it.
>>
> Thanks, Lily. Could you post the stack trace you are seeing on 
> DERBY-5030? It surprises me that it is the same stack trace described 
> in that bug because generateStatistics() no longer calls 
> updateIndexStatsMinion() directly.

Hi Rick,

I've once seen another NPE (Solaris with Java SE 6), unfortunately I've 
lost the stack trace. However, I saw exceptions in both the user thread 
and the istat thread.
I think the istat thread got an NPE in 
DataDictionaryImpl.disableIndexStatsRefresher due to a timing bug 
between error handling in the daemon and the shutdown process. This bug 
is most certainly intermittent, and happens when the data dictionary 
module is stopped before the daemon tries to disable itself due to a 
fatal error (caused by the shutdown).
The NPE itself is simple to avoid, I just need to figure out if 
something else should be done.

I haven't seen DERBY-5030 after the patch for DERBY-5037 went in, but 
I've only tested on Windows Vista and Solaris.

Regarding the OOME, how large is the heap in the nightly runs?


Regards,
-- 
Kristian

>
> Thanks,
> -Rick


Mime
View raw message