You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@chemistry.apache.org by Jukka Zitting <ju...@gmail.com> on 2010/03/18 17:45:16 UTC
OpenCMIS Hudson updates
Hi,
I changed the OpenCMIS Hudson build job to poll svn only once an hour
instead of once every minute to respect the polling guidelines set by
Apache infra.
I also noticed that the job is not configured to notify chemistry-dev@
when a build fails. Should we change that?
PS. I'm trying to figure out a reason for the earlier build timeouts.
BR,
Jukka Zitting
Re: OpenCMIS Hudson updates
Posted by Jukka Zitting <ju...@gmail.com>.
Hi,
On Tue, Mar 23, 2010 at 2:59 PM, Klevenz, Stephan
<st...@sap.com> wrote:
> Hopefully this is not too expensive to run on Apache Hudson landscape.
The dedicated build slaves should be able to handle that without
problems. If needed, we can even tie the job to the vesta.apache.org
slave that only has one executor configured, and thus acts as a
dedicated server for the currently executing build.
The problem is the Hudson master that's consuming about 2GB just to
manage all the build jobs we have at Apache. We're currently trying to
move as much of the build load out to the build slaves as possible to
free up capacity on the master. See builds@ for more details.
BR,
Jukka Zitting
RE: OpenCMIS Hudson updates
Posted by "Klevenz, Stephan" <st...@sap.com>.
Hi Jukka,
> I wonder if this has something to do with the current
> capacity problems we've been seeing on the
> Hudson master.
A resource problem is very likely. The test runs of OpenCMIS are very expensive and I had to adapt MAVEN_OPTS several times because of OutOfMemory error occurs. Current settings are:
-Xms256m -Xmx1024m -XX:PermSize=128m -XX:MaxPermSize=256m
In detail OpenCMIS testing is starting a Jetty server which runs the InMemory repository. Then OpenCMIS runtime junit tests executing through the whole OpenCMIS stack: client runtime, provider layer, network binding until each test case hits finally the jetty server. All of this is quite effortful but ensures a high code quality.
Hopefully this is not too expensive to run on Apache Hudson landscape.
Regards,
Stephan
-----Original Message-----
From: Jukka Zitting [mailto:jukka.zitting@gmail.com]
Sent: Dienstag, 23. März 2010 12:16
To: chemistry-dev@incubator.apache.org
Subject: Re: OpenCMIS Hudson updates
Hi,
On Thu, Mar 18, 2010 at 6:23 PM, Klevenz, Stephan
<st...@sap.com> wrote:
> I'm also analyzing these build problems (timeout, etc.) but until now with no
> success. Sometimes the build hangs and repeating the build without code
> change between works or vice versa. I run a Hudson build locally and in parallel
> that does not show these kind of problems (except real build errors).
There seems to again be a build stuck. I wonder if this has something
to do with the current capacity problems we've been seeing on the
Hudson master. For now I've tied the OpenCMIS build to the two Ubuntu
build slaves, which should avoid problems caused by load on the Hudson
master. Let's see whether that helps.
> I suggest to change back to mailing list notifications after the reason is found
> and build is more stable.
Agreed.
BR,
Jukka Zitting
Re: OpenCMIS Hudson updates
Posted by Jukka Zitting <ju...@gmail.com>.
Hi,
On Thu, Mar 18, 2010 at 6:23 PM, Klevenz, Stephan
<st...@sap.com> wrote:
> I'm also analyzing these build problems (timeout, etc.) but until now with no
> success. Sometimes the build hangs and repeating the build without code
> change between works or vice versa. I run a Hudson build locally and in parallel
> that does not show these kind of problems (except real build errors).
There seems to again be a build stuck. I wonder if this has something
to do with the current capacity problems we've been seeing on the
Hudson master. For now I've tied the OpenCMIS build to the two Ubuntu
build slaves, which should avoid problems caused by load on the Hudson
master. Let's see whether that helps.
> I suggest to change back to mailing list notifications after the reason is found
> and build is more stable.
Agreed.
BR,
Jukka Zitting
RE: OpenCMIS Hudson updates
Posted by "Klevenz, Stephan" <st...@sap.com>.
Hi Jukka,
I have removed the notifications to chemistry-dev because of the many instable builds on Apache Hudson. They had flood the mailing list.
I'm also analyzing these build problems (timeout, etc.) but until now with no success. Sometimes the build hangs and repeating the build without code change between works or vice versa. I run a Hudson build locally and in parallel that does not show these kind of problems (except real build errors).
I suggest to change back to mailing list notifications after the reason is found and build is more stable.
Regards,
Stephan
-----Original Message-----
From: Jukka Zitting [mailto:jukka.zitting@gmail.com]
Sent: Donnerstag, 18. März 2010 17:45
To: chemistry-dev@incubator.apache.org
Subject: OpenCMIS Hudson updates
Hi,
I changed the OpenCMIS Hudson build job to poll svn only once an hour
instead of once every minute to respect the polling guidelines set by
Apache infra.
I also noticed that the job is not configured to notify chemistry-dev@
when a build fails. Should we change that?
PS. I'm trying to figure out a reason for the earlier build timeouts.
BR,
Jukka Zitting