We have had an all-clouds.yaml file that was not being managed on disk
by puppet. Actually apply it to disk so that the template ends up on the
puppetmaster as expected.
Change-Id: I0136cab7c03b1932be5b24ff2e93ea8adb84c20d
Infra uses in all its clouds two projects, one for long-lived
servers (openstackci) and another one for nodepool (openstackjenkins).
Putting this on infracloud controller manifest to automate that.
Change-Id: I75725e9256172058e3d6a7354de064da7f898ce8
Update the system-config manifest to support the simplifying changes
made in the puppet-infracloud module.
This patch will require updates to hiera. We need keys
hpuswest_ssl_cert_file_contents and ssl_key_file_contents added, and
hpuswest_ssl_cert_file_contents must be in the 'infracloud' hiera group
since it is shared to the compute nodes.
Change-Id: I39c70b1077e8b467e0a7e123a694d037ffc77f7a
Depends-On: Ibeea608e965e58c496a95b2f02a4bf6b13e15f0e
We started collecting these things... and never stopped.
exmaple:
find /var/lib/puppet/reports/zm08.openstack.org/ -mtime +5 | wc -l
690
Change-Id: I72dc2bb32c76ae8f2ebd22801e8d3e9924c25d4d
We cannot use default guest/guest as it has been disabled
on our rabbit. So we need to generate a pass in hiera for that,
and pass it properly to the manifest.
Change-Id: I407119383b232f07888dc9821771f1ece383a431
Depends-On: I9582d68ca93f2f7b5742523e273ebf6b5a9c0c13
We can't colocation debian and ubuntu - so allow for each having their
own pool independent pool structure.
Change-Id: I9fe0fa9769a5c0f6a852844b2db0dfb620e91375
We need the gpg signing keys for the upstream repos in root's keyring.
Fuel uses golja-gnupg for importing keys - so seems like a good thing
for us to use too.
Change-Id: I343be9397ba53f494485d1ff0f1e1e4c96c0273f
In the current setup, the reprepro db and lists dirs are being written
into /root. Move them into AFS so that we store all the data in the same
place.
Also, update the script to point to the confdir for reprepro commands.
Note: This will require a manual move of the current db and lists dirs
from /root to /afs/.openstack.org/mirror/apt before this is run.
Change-Id: I2638621e6af98296704e680227fae174d8a0e541
Our mirror vhost config uses symlinks in /var/www to control what is
published. Add an entry for the apt repo.
Change-Id: I301c61282e53d4cf596efe45b27dadce797f947e
Since these are baremetal hosts, they need to come from a static
inventory not the openstack inventory. Fortunately, thats pretty easy.
Also setting infracloud groups to be children of disabled to keep them
disabled until we are ready.
Change-Id: I87ed4008ed9c4867f79bbb5fbb6be53707b42625
The reprepro class in this is in-tree rather than in its own module
purely for ease of getting started. It's also highly hard-coded rather
than flexible.
This change will need a mirror.apt volume and service/reprepro principal
and keytab to be created before it lands.
Allow for pool trimming after a 2 hour delay. Each devstack run of
apt-get update should be able to be assumed to be valid for the length
of the devstack. For that reason, only delete files that are
unreferenced during the subsequent mirror run, ensuring at least a 2
hour delay between becoming unreferenced and going away.
Local testing indicates that a trusty mirror is 86G.
Change-Id: I84f6a0391f80e6bf567c4bfc18a41bd270fe8c01
Upstream Gerrit confirms that the default value for the httpd.maxQueued
value was set too low. The fix[1] for it was to change the default to
200 however the fix is only in version 2.12+ not in 2.11.
clarkb confirmed that we are seeing the same error (stack trace linked
in 1st ml thread in upstream fix[1]) in review.o.o logs. This likey
indicates that VM starvation (due to GC) is making Gerrit to slow to
process incoming client requests so it pushes more requests into the
queue than it can handle. Once the queue is exhausted the requests
get rejected with 500 errors. Increasing the max number of client
connections may help alleviate this situation until Gerrit can catch
up.
[1] https://gerrit-review.googlesource.com/#/c/70627
depends-on: Iea202fcfe7af4a9d5b4c3a360124c24649b1133b
Change-Id: I8d6c8d9ec6348fc6b950744237e527133a625bbe
This adds a jenkins-owned working directory for the wheel slave.
Actual work will occur in subdirectories that can be easily deleted.
Change-Id: Ie91e7e001c4a45fe70abca7a18671a0580d19d9f
This makes possible to apply openstack_project::slave_common class
to debian in chroot, e.g. to create a debian image with puppet
element using diskimage-builder
Change-Id: I7a661c8b5d4bb632eb32b5018c55474c1c66a4e8
This patch adds AFS, and the admin keytab, to our release slave,
so that we may execute `vos release` when a mirror update
is requested.
Change-Id: I9c8531304fb74b9bafca85d58b5e9c1c0bc21f6a
This patch adds a vhost that exposes /afs/openstack.org/mirror/wheel
to the internet under <hostname>/wheel. It also adds appropriate
rewrite rules.
Change-Id: Id514a76f7759b663b2ea3a3ec4aa2dfe083d6666
This adds a hiera-loaded keytab for the wheel slaves. It may
be used when writing wheels to the AFS share directory.
Note: This will require the creation of a keytab for the
wheel mirror, as well as appropriate folders and access rights
on AFS.
Change-Id: I81c39d420d8ac70def57949ea0d4c323b8797086
