Check jobs failed due wrong drush make execution. This fix
now build up drupal distribution from installation profile
and contrib module parts instead of directly checking out
the code from git repository HEAD.
Change-Id: I77d053ac7b8778d7f0abc0cbe6b4a5ce62affd63
Closes-Bug: #1331109
as we're currently at quota, with about 1.5x quota worth of load
in the check queue. That's put us in a state that f20 nodes are
not being allocated at all. Which means that devstack and devstack-gate
changes are currently stalled.
The fix for the top gate bug is probably a devstack fix. Which puts us
in a position where we can't make things better because we're
stuck waiting for f20 nodes.
Move to experimental for now. We need a longer term fix to prevent
us from getting into this situation before moving it back to check.
Change-Id: I7f44cb9896f646060d45ac571ddb000b721812a6
We need to start doing this before we change grenade to not default
to spitting out everything to the console.
Change-Id: I58d32c27e78d2d6eb791f2d29c3aa886a5218c68
On centos, pip install goes into /usr/bin, not /usr/local/bin.
This means that explicitly calling out /usr/local/bin/run-mirror
is just directly wrong there.
Change-Id: Ieb088a75650673c3a912025ab507d4cfc253e5ec
Add elasticsearch07 node. Move the elasticsearch discover node to
elasticsearch02 instead of 01 as we are moving away from 01 as part of
the 07 addition.
Change-Id: I2aa857ec4984ae1fc2f8e27f437f8ecc61d24fbd
The large-ops test is there to prevent massive performance regressions,
and we are ratcheting up this test over time. We don't want to have to
maintain a separate large-ops job for stable/havana especially since
stable releases only accept 'safe' backports from master. So if a patch
is being considered for backport it already has passed master large-ops
and a human reviewer evaluating the patch explicitly for how safe it is
to backport.
The new large-ops number broke this test for stable/havana, so turn off
the job on stable/havana but keep it on stable/icehouse as its still
working there.
Closes-Bug: #1329042
Change-Id: I5fb64f8537dfffca79170000310a02353b7d1b0e
Our Elasticsearch nodes need cinder volumes attached to them prior to
being puppeted. Make the regex for the elasticsearch nodes in site.pp
match only existing nodes so that new nodes can be created and have
their volumes attached before proper puppeting.
Change-Id: If12c67174315b8480998148e1325de59ecdba372
A side effect of fully symmetric testing of the olsotest module is
that it will end up joining the world into a single gating queue
due to transitive dependencies. It already put ironic in line with
the rest of the gate because the following cross testing:
ironic <-> dib <-> oslotest <-> oslo.messaging <-> all of openstack
As stackforge projects included oslotest, we might have gotten
similar things.
Here is the proposed solution to get us out of this:
* remove oslotest tests from the gate queue of projects that use it
* continue to run oslotest tests on check queues for those projects
(this is basically 'gate lite', as clean check still makes them
have to pass these tests)
* leave oslotest gate tests on oslotest. This handles the case where
a series of olsotest patches merge conflict badly, but can't get
caught until the gate.
This means there is an exposure where "just the right" set of patch
lands in a project which break oslotest usage could cause a temporary
wedge of that project + oslotest. However I think that's a localized
enough problem we should take that risk and manage it instead of
the current monolythic pipeline problem.
Change-Id: I371ddf7c1de42d82773468064b5706ae5c36ad8b
This job is passing only 66% of the time in the gate, 30 fails in
the last 48hrs. That's too low to be allowed to vote in the gate.
Due to the oslotest overlap, failing on this configuration now
impacts all of OpenStack. Some initial analysis shows that this
configuration is doing *a lot* more interaction with the network
apt servers. That's something which fundamentally needs to be
limitted.
Instead this job is going to need to work on building a new nodepool
image that includes the cloud archive, instead of installing it
on the fly.
This leaves the check job voting, but will not impact the rest of
OpenStack when ironic or diskimagebuilder fail in the gate.
Change-Id: I2652b777e4e24034f715dfecb71a87f128f14644
Python logging has a WatchedFileHandler which will do the correct thing
for logrotate when copytruncate is not used. Do this to avoid having
multiple copies of the same log in place. Also, this should avoid a race
where log data is lost between copying and truncating done by logrotate.
Change-Id: Idbbc7869004a03b8f19cfbd81ad07f5584ce1964
Zuul must be explicitly told in its layout file which jobs should have
swift parameters passed in. Start with experimental-swift-logs.* passing
in the swift creds name as 'logs'. This matches up with the existing
config-pep8 jobs we have in the experimental queue which is being used
for testing.
Change-Id: I4816befb67bd3c793a72b8fcd403366cb9a523d4
Adjust nodepool limits for rax-dfw, rax-iad and rax-ord to reflect
our current reality. Our primary quota constraints there are
maxTotalInstances and maxTotalRAMSize, the latter needing to be
divided by 8192 to get an upper bound on possible node count. Take
whichever is lower in each region, subtract the number of images
configured for it to make room for image update template instances
while running at capacity, and knock off two more for breathing
room.
The drastic reduction in max-servers for rax-iad is due to a recent
and so far unexplained drop in maxTotalRAMSize there, which is being
investigated separately and can be readjusted upward once the cause
is addressed.
Change-Id: Iec601ed87bae9a048525ebcde37deb373b688f4d
We attempt to set hostname in /etc/hosts if the hostname does not exist
there. However when the $HOSTNAME variable is empty we grep for
/etc/hosts from stdin which has no data. Check that $HOSTNAME is not
empty before we grep for it.
Change-Id: Ie35fd997f2f569ebea6614f91609281a991f8d6a
The OpenStack Training Guides team has been sheltering under
the loving arms of the openstack-manuals repo for the recent past.
It's now time to move them out to further their independent work.
This change adds a project:
openstack/training-guides
seeded with the training-guides subdirectory from openstack-manuals,
and attempts to configure the jenkins/zuul jobs so that the existing
docbook check, gate and post will also run on the new project.
Change-Id: I1aed9efa8ca3fd106872cec248cd7c4964fa9adf
We are making this and other CLA enforcement groups public, since
they are not sensitive information (lists of committers are already
public and we used to keep these tracked in a wiki article before
enforcement within Gerrit), and since there is value in being able
to query this status in a read-only manner to aid in building
contributor analyses.
Change-Id: I459a066612fe014790086ac04cbeb5eebd6e062d
This gets us closer to the behavior we had prior to upgrading from
2.4 to 2.8, and is generally considered desirable since we want
people to be able to review and vote on commit message contents.
Change-Id: I7051a606f1d501e5aef6510d42ef928456d07303