db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Knut Anders Hatlen <knut.hat...@oracle.com>
Subject Re: Jenkins build became unstable: Derby-10.9-suites.All #84
Date Tue, 28 May 2013 08:59:41 GMT
Knut Anders Hatlen <knut.hatlen@oracle.com> writes:

> Dag Wanvik <dag.wanvik@oracle.com> writes:
>> Is this a real regression, or is some server hanging on the Jenkins
>> server?
> Looks like these two builds started at about the same time and on the
> same slave node:
> https://builds.apache.org/view/A-D/view/Derby/job/Derby-10.10-suites.All/21/
> https://builds.apache.org/view/A-D/view/Derby/job/Derby-10.9-suites.All/84/
> And both builds had failures in network tests.
> I'll see if I can change the configuration so that they either don't run
> at the same time, or at the same node, or use the same set of ports.

I ended up with this scheme for now (had to back-port DERBY-6078 to 10.9

Derby-10.10-suites.All is set up to run with derby.tests.basePort=1530,
and Derby-10.9-suites.All with derby.tests.basePort=1560. This way, they
won't conflict with each other or with the derbyall jobs when running on
the same node.

We cannot configure the port used by derbyall, so I changed the poll
interval instead. Both Derby-10.9-derbyall and Derby-10.10-derbyall used
to check the Subversion repository for updates at midnight (in some time
zone). Now 10.9 is checked at midnight and 10.10 at midday. That will
hopefully bring them far enough apart so that they don't overlap, even
if they happen to run on the same node.

Knut Anders

View raw message