Re: hammer 0.94.10 QE status

[Gregory Farnum <gfarnum@redhat.com>]

On Thu, Feb 16, 2017 at 9:06 PM, Nathan Cutler <ncutler@suse.cz> wrote:
> From re-reading this thread I determined that the only two outstanding
> issues are upgrade/firefly-x and rgw+centos (can anyone confirm?)
>
> I don't think upgrade/firefly-x should be a blocker. Rationale: as of the
> jewel release, firefly clusters were supposed to be upgraded to at least
> hammer. As of the kraken release, support of hammer has been focused on
> facilitating users to upgrade from hammer to jewel. In other words, firefly
> clusters should already have been upgraded and upgrading from firefly to
> hammer is no longer supported.

Woah, that can't be the case. Just because we're not supporting
firefly by putting out new releases doesn't mean it's acceptable to
prevent users from upgrading it to newer LTSes!

I've no idea what's broken there, so eg more valgrind failures are not
a big deal, but if the upgrade itself is failing in some way that's
going to be a problem. o_O
-Greg

>
> As for the rgw valgrind/libtcmalloc failures on centos, I made a
> wip-hammer-baseline branch based on "hammer" but with a different SHA1 ("git
> commit --amend --reset-author") and pushed it to ceph/ceph-ci.git so Shaman
> will re-build it. This should ensure that "flavor=notcmalloc" really means
> what it says. Re-running rgw+centos on this wip-hammer-baseline branch will
> tell us more.
>