You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@jena.apache.org by Apache Jenkins Server <je...@builds.apache.org> on 2011/10/17 02:08:29 UTC

Jenkins build became unstable: Jena_TDB_Nightly #9

See <https://builds.apache.org/job/Jena_TDB_Nightly/9/changes>



Re: Build problems

Posted by Paolo Castagna <ca...@googlemail.com>.
Hi Andy

Andy Seaborne wrote:
> This is unstable because it run on Ubuntu1; in #8 it ran on Ubuntu2 and
> passed.

I am try to think what the cause could be...

A VM?

... and something go wrong when we try to spot which OS tests are running on?

> Is there anyway we can delete the state of everything and start again?
> At the moment, CI fails quite frequently for no reason (i.e. when there
> is no change to the codebase, and it fails in a deterministic test)
> 
> Restricting it to Ubuntu2 seems pointless - a problem that will
> resurface somewhere else.

Agree.

> 
> It's already set to "Always check out a fresh copy"
> 
> (I've reordered the tests just for the sake of making a change as well)
> 
> I don't get this error when running locally nor when I building for
> Apache (mvn2 or mvn3).
> 
> -----
> 
> ARQ Nightly #4 had been running for 8 hours (on Ubuntu2, so that machine
> wasn't free for the TDB/Nightly job).

Each job is configured to fail if stuck for more 30 minutes.

> ARQ Nightly has an error "pom2.xml" no such file, but I see pom2.xml in
> SVN.
> 
> I restarted it and it succeeded.
> 
> -----
> Do we need to run a nightly build if there has been no changes (e.g. IRI)?

No. The aim is to publish SNAPSHOTs as soon as there is a change in SVN.
So, that the SNAPSHOTs in the Apache Maven repository are always up to date.

I'll rename "nighly" to "snapshot" and instead of schedule them every night
I make them check for SVN updates.

When pom2.xml files go away, there will be no need for these additional jobs.
We will have only one job per module (and that could publish SNAPSHOTs when
code changes.

Paolo

> 
> 
>     Andy
> 
> 
> On 17/10/11 01:08, Apache Jenkins Server wrote:
> 
> Subject: Jenkins build became unstable:  Jena_TDB_Nightly #9
> 
> See<https://builds.apache.org/job/Jena_TDB_Nightly/9/changes>
> 

Re: Build problems

Posted by Andy Seaborne <an...@apache.org>.
On 17/10/11 16:52, Paolo Castagna wrote:
>
>
> Andy Seaborne wrote:
>> This is unstable because it run on Ubuntu1; in #8 it ran on Ubuntu2 and
>> passed.
>>
>> Is there anyway we can delete the state of everything and start again?
>> At the moment, CI fails quite frequently for no reason (i.e. when there
>> is no change to the codebase, and it fails in a deterministic test)
>>
>> Restricting it to Ubuntu2 seems pointless - a problem that will
>> resurface somewhere else.
>>
>> It's already set to "Always check out a fresh copy"
>>
>> (I've reordered the tests just for the sake of making a change as well)
>>
>> I don't get this error when running locally nor when I building for
>> Apache (mvn2 or mvn3).
>
> ConfigTest.getTestingDirDB() is "tmp/testing/DB".
> init() creates those directories only once, if the static boolean initialized is false.
> TestTransRestart deletes that directory in the cleanup() method.
> TestTransRestart setup() method tries to recreate it via FileOps.ensureDir(path) which uses File.mkdir() (not mkdirs).
> Can this be the cause of the problems?

I don't think so - it would fail always if this were the case.

@Before does:

     cleanup() ;
     FileOps.ensureDir(path) ;

so the directory is deleted, then FileOps.ensureDir is called anyway and 
the directory recreated, fresh and empty.

Also the directory is used by other tests in the suite.

Could try creating/ensuring it once @BeforeClass and doing a clean.

> It does not account for why does not fail on ubuntu2|4 or locally though.
>
> Could it be a file permissions issue?

It is used by other tests in the suite and I'd expect deterministic 
failures on all "ubuntu" nodes because the jobs delete everything and 
check out again.

>
> Using "/tmp" instead of a temporary directory in the workspace (i.e. "tmp") seems to solve this problem.

i.e. a fresh-per-run directory.
(minor but by keeping the workspace in the project is gets left behind 
in case of real failures (this is when developing))

What about using

ConfigTest.getTestingDirUnique()

but if "/tmp" works, leave it (I am just suspicious it will reappear).

	Andy

>
> Paolo


Re: Build problems

Posted by Paolo Castagna <ca...@googlemail.com>.

Andy Seaborne wrote:
> This is unstable because it run on Ubuntu1; in #8 it ran on Ubuntu2 and
> passed.
> 
> Is there anyway we can delete the state of everything and start again?
> At the moment, CI fails quite frequently for no reason (i.e. when there
> is no change to the codebase, and it fails in a deterministic test)
> 
> Restricting it to Ubuntu2 seems pointless - a problem that will
> resurface somewhere else.
> 
> It's already set to "Always check out a fresh copy"
> 
> (I've reordered the tests just for the sake of making a change as well)
> 
> I don't get this error when running locally nor when I building for
> Apache (mvn2 or mvn3).

ConfigTest.getTestingDirDB() is "tmp/testing/DB".
init() creates those directories only once, if the static boolean initialized is false.
TestTransRestart deletes that directory in the cleanup() method.
TestTransRestart setup() method tries to recreate it via FileOps.ensureDir(path) which uses File.mkdir() (not mkdirs).
Can this be the cause of the problems?
It does not account for why does not fail on ubuntu2|4 or locally though.

Could it be a file permissions issue?

Using "/tmp" instead of a temporary directory in the workspace (i.e. "tmp") seems to solve this problem.

Paolo

Build problems

Posted by Andy Seaborne <an...@apache.org>.
This is unstable because it run on Ubuntu1; in #8 it ran on Ubuntu2 and 
passed.

Is there anyway we can delete the state of everything and start again? 
At the moment, CI fails quite frequently for no reason (i.e. when there 
is no change to the codebase, and it fails in a deterministic test)

Restricting it to Ubuntu2 seems pointless - a problem that will 
resurface somewhere else.

It's already set to "Always check out a fresh copy"

(I've reordered the tests just for the sake of making a change as well)

I don't get this error when running locally nor when I building for 
Apache (mvn2 or mvn3).

-----

ARQ Nightly #4 had been running for 8 hours (on Ubuntu2, so that machine 
wasn't free for the TDB/Nightly job).

ARQ Nightly has an error "pom2.xml" no such file, but I see pom2.xml in SVN.

I restarted it and it succeeded.

-----
Do we need to run a nightly build if there has been no changes (e.g. IRI)?


	Andy


On 17/10/11 01:08, Apache Jenkins Server wrote:

Subject: Jenkins build became unstable:  Jena_TDB_Nightly #9

See<https://builds.apache.org/job/Jena_TDB_Nightly/9/changes>