patch process failures

[Nigel Daley <nd...@yahoo-inc.com>]

Some of you have asked why the patch process is failing so  
frequently.  Our unit test suite seems to have destabalized again.

FYI, when you go to a link from the Jira comment like this:
   <http://lucene.zones.apache.org:8080/hudson/job/Hadoop-Patch/430/ 
testReport/>
and see entries in the "Failed Tests" section like this:
   org.apache.tools.ant.taskdefs.optional.junit.JUnitTask$1.unknown
it means that a test took longer than 15 minutes to run and was  
therefore killed (timeout).
To find out which one, you have to go to the Console Output ([raw]  
link) and search for
FAILED (console output may take a while to load).

I'm seeing TestTable and TestTimestamp timeout most frequently.  Also  
TestSymLink, TestDecommission, and TestDFSUpgrade fail occasionally.


hth,
Nige

Re: patch process failures

[Nigel Daley <nd...@yahoo-inc.com>]

On Jul 18, 2007, at 10:15 AM, Michael Stack wrote:

> Thanks for the info below Nigel.  Helps.
>
> Hopefully HADOOP-1616 committed yesterday will make a difference  
> regards TestTable sometimes hanging.

Great!  Can you make sure this made it into the 0.14 branch.  If not,  
I think it should be merged into the branch, with the necessary Jira  
updates.

> But whats is odd is that the two hbase tests cited actually run for  
> a short time on my test bench but they do run toward the end of the  
> hbase suite.  Could it be that the 15 minutes timeout is for all  
> hbase tests?

No, the timeout is passed to junit where it is used per test.

Cheers,
Nige

> Nigel Daley wrote:
>> Some of you have asked why the patch process is failing so  
>> frequently.  Our unit test suite seems to have destabalized again.
>>
>> FYI, when you go to a link from the Jira comment like this:
>>   <http://lucene.zones.apache.org:8080/hudson/job/Hadoop-Patch/430/ 
>> testReport/>
>> and see entries in the "Failed Tests" section like this:
>>   org.apache.tools.ant.taskdefs.optional.junit.JUnitTask$1.unknown
>> it means that a test took longer than 15 minutes to run and was  
>> therefore killed (timeout).
>> To find out which one, you have to go to the Console Output ([raw]  
>> link) and search for
>> FAILED (console output may take a while to load).
>>
>> I'm seeing TestTable and TestTimestamp timeout most frequently.   
>> Also TestSymLink, TestDecommission, and TestDFSUpgrade fail  
>> occasionally.
>>
>>
>> hth,
>> Nige
>

Re: patch process failures

[Michael Stack <st...@duboce.net>]

Thanks for the info below Nigel.  Helps.

Hopefully HADOOP-1616 committed yesterday will make a difference regards 
TestTable sometimes hanging.

But whats is odd is that the two hbase tests cited actually run for a 
short time on my test bench but they do run toward the end of the hbase 
suite.  Could it be that the 15 minutes timeout is for all hbase tests?

Thanks,
St.Ack

Nigel Daley wrote:
> Some of you have asked why the patch process is failing so 
> frequently.  Our unit test suite seems to have destabalized again.
>
> FYI, when you go to a link from the Jira comment like this:
>   
> <http://lucene.zones.apache.org:8080/hudson/job/Hadoop-Patch/430/testReport/> 
>
> and see entries in the "Failed Tests" section like this:
>   org.apache.tools.ant.taskdefs.optional.junit.JUnitTask$1.unknown
> it means that a test took longer than 15 minutes to run and was 
> therefore killed (timeout).
> To find out which one, you have to go to the Console Output ([raw] 
> link) and search for
> FAILED (console output may take a while to load).
>
> I'm seeing TestTable and TestTimestamp timeout most frequently.  Also 
> TestSymLink, TestDecommission, and TestDFSUpgrade fail occasionally.
>
>
> hth,
> Nige