[Hawkular-dev] Hawkular project switching to Docker-based Travis workers; was: Hitting Travis Limits

[Peter Palaga]

In my case, switching to Travis container-based infrastructure [1][2] 
helped - see [3] for how to do that.

Hard to say which particular aspect of the container-based Travis worker 
has helped - here are some possible candidates:
* more RAM
* more CPUs (2 vs. 1.5)
* faster network -> faster artifact downloads and therefore shorter
   total build times -> less probable job timeouts

Another important pro for switching to docker workers is that the job 
starts much faster than with the traditional VM worker.

-- P

[1] http://docs.travis-ci.com/user/workers/container-based-infrastructure/
[2] 
http://blog.travis-ci.com/2014-12-17-faster-builds-with-container-based-infrastructure/

[3] 
https://github.com/ppalaga/hawkular/commit/351e046c3d383023e5a35fe5c15891f45c8b1c28#diff-354f30a63fb0907d4ad57269548329e3R4



On 2015-05-27 18:06, Thomas Segismont wrote:
> Le 27/05/2015 17:02, Heiko W.Rupp a écrit :
>>> I am not sure myself, which of the two is my case. Before I spend any
>>> time on that, has anybody ever succeeded to start wildfly inside Travis
>>> and run some integration test?
>>
>> We had Hawkular-metrics running on Travis including tests (iirc both
>> with memory and Cassandra backend).
>
> Well, the memory backend is long gone. But as a matter of fact, the
> Metrics build starts a Wildfly server twice: first in the rest-tests
> module, then in the ptrans module.
>
>
>
> _______________________________________________
> hawkular-dev mailing list
> hawkular-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/hawkular-dev
>