On Tue, Mar 13, 2018 at 6:50 PM, Martin Sivak <msivak@redhat.com> wrote:
Hi,just two or three notes from me:- I had to find a suitable hardware first- we spent quite some time to actually get OST up and running (missing CPU flags issue in my case)- he-basic-ansible-suite-4.2 failed because of package dependencies (imageio, but probably related to what Sandro already mentioned)- he-basic-suite-master fails for me because of the SDK version, but I have not way to check which version is requiredJust the last three issues cost me almost the whole day of playing with lago and OST instead of actually thinking about improvements.The contribution guide would help indeed as would better compatibility and some better way of dependency management.Btw, what would it take to allow ansible based tests? It should not be too hard to generate an inventory file for the deployment and launch all playbooks in the test directory. This would make contribution of simple tests even easier I think.Best regardsMartin SivakOn Tue, Mar 13, 2018 at 5:26 PM, Sandro Bonazzola <sbonazzo@redhat.com> wrote:______________________________4 people accepted calendar invite:- Devin A. Bougie- Francesco Romani- Jiri Belka- suporte, logicworks4 people tentatively accepted calendar invite:- Amnon Maimon- Andreas Bleischwitz- Arnaud Lauriou- Stephen Pesini2 mailing lists accepted calendar invite: users@ovirt.org, devel@ovirt.org (don't ask me how) so I may have missed someone in above list4 patches got merged:13 patches has been pushed / reviewed / rebasedFeedback from the event:- "if we want to add many more tests to OST, and I think we do, we need to do some change there to allow that. Current framework is simply not scalable enough"- not joining the hackathon because "I'd be like an elephant in a porcelain shop"- "I'm not sure I'm OK with the flood of suites that we have - the more we have, the harder it is to sync and maintain but more importantly - to run."- "We can't keep adding new suite for each parameter we want to test, it adds overhead to monitoring, resources and maintenance."- invite wasn't clear enough. I found people on #ovirt on Freenode and on Red Hat IRC servers and redirected them to OFTC IRC server (my fault, hopefully managed to workaround it by talking to people)Lessons learned:- Calendar invites to mailing lists doesn't work well, need a different way to track mailing list members joining the events.- Invites needs to be pedantic on how to join the event, not leaving space for interpretation and misunderstanding.- We need a contribution guide to ovirt-system-test: we need to make people comfortable in trying to add a new test and we need to ensure that we won't reject a day of work because the patch doesn't match core contributors plannings on number of suites, resources and so on- The ovirt-system-tests check patch script is not good enough. It triggers too many sequential suites on every single patch pushed, and fails due to timeout taking more than 6 hours to complete.- The way ovirt-system-test collects rpms from defined repos is not smart enough: it doesn't take the latest version of a given package, just the first found in sequential order of the repos,Thanks everyone who participated to the event! if you have time please continue improving ovirt-system-test even if today event is almost completed!--SANDRO BONAZZOLA
ASSOCIATE MANAGER, SOFTWARE ENGINEERING, EMEA ENG VIRTUALIZATION R&D
_________________
Devel mailing list
Devel@ovirt.org
http://lists.ovirt.org/mailman/listinfo/devel
_______________________________________________
Infra mailing list
Infra@ovirt.org
http://lists.ovirt.org/mailman/listinfo/infra