On Thu, Feb 16, 2017 at 3:35 PM, Sage Weil <sweil@xxxxxxxxxx> wrote: > On Thu, 16 Feb 2017, John Spray wrote: >> On Thu, Feb 16, 2017 at 11:14 PM, Yuri Weinstein <yweinste@xxxxxxxxxx> wrote: >> > I can't say I understand details on what you want to achieve, but I'd >> > try combination of "--subset" and "--filter" >> > >> > I think/hope "--filter" would work on whatever you get out of "--subset" filter. >> >> Ah, the key thing is that I don't want to just run a filtered part of >> the sample, I want to control which aspects of the test are sampled, >> vs which are all run. I am also having trouble putting together the >> right words to explain this :-) >> >> So if I've got a suite that looks something like this: >> >> mysuite/rados_pool/ec.yaml >> mysuite/rados_pool/replicated.yaml >> mysuite/rados_store/bluestore.yaml >> mysuite/rados_store/xfs.yaml >> mysuite/cephfs_inline/yes.yaml >> mysuite/cephfs_inline/no.yaml >> mysuite/tasks/myworkunit.yaml >> >> Then I want to always include all permutations of cephfs_inline, while >> only including a subset of the permutations of rados_pool and >> rados_store. There are 4 permutations of the rados_* bits, so I'd be >> passing something like --subset=1/4, but for it to work the way I want >> I need a way to tell teuthology-suite to only do the subset part on >> the rados bits, and leave the cephfs bits alone. I'd love that. >From QE standpoint it'd be valuable if we can run suites with jobs tagged "nightlies" and then maybe tagged "all" and maybe "point_release", etc. as we define tags. It would also allow us to manage sepia lab load better ! > > +1 > > This would be super useful, IMO! > > My main question is whether it's something we might want to encode into > the suite itself, so that certain permutations sample before others do. > I'm not quite sure how to put it... maybe it's simplest to say that some > permutations would *always* sample. There are lots of cases where I want > to mix in additional variables but I don't expect them to corelate > with anything else, but end up blowing up the test count anyway. > > sage > > >> >> John >> >> >> >> >> >> >> > On Thu, Feb 16, 2017 at 3:11 PM, John Spray <jspray@xxxxxxxxxx> wrote: >> >> Evening all, >> >> >> >> I'm wondering if this is something that's buried in the existing code, >> >> or if others have thought about it -- I'd like to be able to do the >> >> --subset sampling, but only on some aspects of a suite. Specifically, >> >> in the fs suite we have some RADOS-level options like xfs vs. >> >> bluestore, forthcoming EC vs. replication, and I'd like to sample >> >> those, while still running all the cephfs-specific permutations >> >> exhaustively. >> >> >> >> So if this doesn't already exist I guess it could look something like >> >> --suite fs --subset=1/3, --subset-filter=rados_* (e.g. if I adopted a >> >> naming convention where dirs/files controlling the rados-ish bits wre >> >> named like that). >> >> >> >> Any thoughts? >> >> >> >> John >> >> -- >> >> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in >> >> the body of a message to majordomo@xxxxxxxxxxxxxxx >> >> More majordomo info at http://vger.kernel.org/majordomo-info.html >> -- >> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in >> the body of a message to majordomo@xxxxxxxxxxxxxxx >> More majordomo info at http://vger.kernel.org/majordomo-info.html >> >> -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
