The py3 battle goes well. A bunch of tests have been pinned to ubuntu nodes (because they have both py2 and py3). Including s3-tests. However... there are a bunch of rgw tests combine s3-tests (needs ubuntu for py2) and vlagrind (needs centos for valgrind notcmalloc build). We could 1- disable these tests 2- add a ubuntu notcmalloc builder and see if valgrind is better in bionic than it was in the xenial and earlier 3- convert s3-tests to py3 already I'm hoping/begging/pleading that someone will take a serious look at s3-tests and tell me how hard it is. I would love to merge the py3 change tomorrow (Tuesday).

Thanks! sage On Mon, 16 Dec 2019, Sage Weil wrote: _______________________________________________ Dev mailing list -- dev(a)ceph.io To unsubscribe send an email to dev-leave(a)ceph.io

On Mon, Dec 16, 2019 at 6:37 PM Sage Weil <sweil(a)redhat.com> wrote: If we want to merge soon, then ignoring these until they are fixed sounds like the best past forward.

Ali mentioned he had these almost done for py3 - I wonder how much is left, I can help out there if needed. _______________________________________________ Dev mailing list -- dev(a)ceph.io To unsubscribe send an email to dev-leave(a)ceph.io

Hey everyone, The python 3 port (https://github.com/ceph/s3-tests/pull/337) is working on Ubuntu for both the s3-tests and the s3-tests-website task. Additionally, the RGW team has decided to remove the readwrite and roundtrip s3-test tests. The only thing left to do is to make sure everything works on Centos 8 in teuthology. Best, Ali On Tue, Dec 17, 2019 at 8:40 AM Casey Bodley <cbodley(a)redhat.com> wrote: > > On 12/16/19 6:41 PM, Alfredo Deza wrote: > > On Mon, Dec 16, 2019 at 6:37 PM Sage Weil <sweil(a)redhat.com> wrote: > >> The py3 battle goes well. A bunch of tests have been pinned to > >> ubuntu nodes (because they have both py2 and py3). Including s3-tests. > >> > >> However... there are a bunch of rgw tests combine s3-tests (needs > ubuntu > >> for py2) and vlagrind (needs centos for valgrind notcmalloc build). We > >> could > >> > >> 1- disable these tests > >> 2- add a ubuntu notcmalloc builder and see if valgrind is better in > bionic > >> than it was in the xenial and earlier > >> 3- convert s3-tests to py3 already > >> > >> I'm hoping/begging/pleading that someone will take a serious look at > >> s3-tests and tell me how hard it is. > >> > >> I would love to merge the py3 change tomorrow (Tuesday). > > If we want to merge soon, then ignoring these until they are fixed > > sounds like the best past forward. > > +1 > > It looks like we'll also need to look into the ragweed and swift tests, > which run in the same rgw/verify jobs with s3tests. Ragweed is pretty > small so it shouldn't take much effort, but our swift tests are from a > 2011 fork of the openstack/swift repo and I don't think anyone's looked > at running those with python3. We'll probably need to pull up our fork, > but that means figuring out how to run the tests (I believe they need > tox instead of nose?), then auditing all the new test failures. That's a > project that we've long put off, and I don't think it's reasonable for > that to block your progress here. > > > > > Ali mentioned he had these almost done for py3 - I wonder how much is > > left, I can help out there if needed. > > > >> Thanks! > >> sage > >> > >> > >> > >> On Mon, 16 Dec 2019, Sage Weil wrote: > >> > >>> I'm working on getting the test suite to run on the py3 branch, > starting > >>> with rados. One of the failures I've run into is that s3-tests runs > with > >>> python 2 instead of 3. This needs to be converted asap in order for > us to > >>> make the jump to python 3. > >>> > >>> I started with a few trivial changes here > >>> > >>> https://github.com/ceph/s3-tests/pull/333 > >>> > >>> but I don't really know what I'm doing. Can someone more familiar > with > >>> s3-tests and/or python take over? > >>> > >>> One question I had was around the s3tests vs s3tests_boto3 > directories... > >>> which is the one that matters, and why is the old one still around? > Or > >>> are they both still used and maintained? > >>> > >>> My other question is how the branches work. I'm guessing we should > >>> actually make the change to teh master branch and then cherry-pick it > to > >>> ceph-master? I'm surprised how far apart master and ceph-master are. > >>> > >>> Thanks! > >>> sage > >>> _______________________________________________ > >>> Dev mailing list -- dev(a)ceph.io > >>> To unsubscribe send an email to dev-leave(a)ceph.io > >>> > >>> > >> _______________________________________________ > >> Dev mailing list -- dev(a)ceph.io > >> To unsubscribe send an email to dev-leave(a)ceph.io > >> > > _______________________________________________ > > Dev mailing list -- dev(a)ceph.io > > To unsubscribe send an email to dev-leave(a)ceph.io > > > _______________________________________________ > Dev mailing list -- dev(a)ceph.io > To unsubscribe send an email to dev-leave(a)ceph.io > >