This commit adds a new nodepool script to use subunit2sql to preseed
testrepository with timing data to use for scheduler optimization.
This uses aggregate data from all the tests stored in the subunit2sql
db from all the test runs to generate a subunit stream with test
execute times which equal the rolling average stored in the tests
table.
This also adds a call out to the new prepare_tempest_testrepository
script in order to pre-seed tempest's testrepository with data from
the subunit2sql DB. This will enable the testr scheduler to perform
some worker balance optimization based on the average run_times from
all the previous gate runs.
Change-Id: I04404a55baf4da2745760246b8606ed3db52c03a
Set up a venv for zuul-swift-logs and install its dependencies
(python-magic) there instead of as system packages.
For DIB use 90 as uploading logs may be considered more important
than installing zuul.
Change-Id: Id0fc01729853e65c81cdd50e4ffa4d0d6de00ae9
Recent versions of pip will create ~/.cache if it doesn't already exist.
If this happens while running "sudo pip" the resulting dir will be owned
by root and 700 - which breaks anything else on the system that wants to
use ~/.cache
I *think* this finds all instances of "sudo pip" in this repo and fixes
them to use -H. This should mean they always run in the right ~
Change-Id: I47ddb3b591df6ac2100f09b38c9b8a03cb1ba1ff
closes-bug: #1405732
This will enable the flexible configuration of the XVA and the URL
through nodepool, thus it's not needed to push an infra change for
changing those values. Also if you do not have those values configured,
sensible defaults are provided.
Change-Id: Ie2be50f842b4220d4e7f9d8e830d3df3262a6353
The --disablerepo flag for yum will fail if the repo being disabled is
not already enabled. While this is not a problem for HP cloud images
that already have epel enabled, it makes it hard for users to use the
script on non-HP cloud nodes. This commit detects whether epel is
enabled and uses the disable flag appropriately.
Change-Id: I26b6fbdc390b99b0e289ac0e35a5e616d5d3682a
We're done with the Puppet 2-to-3 migration, so we no longer need to
explicitly override the puppet version to 2 when bootstrapping.
Change-Id: Ib04e8dda5050ae0138fba248ff28f35819fdf254
dib has a new declarative package installation syntax which allows
specifying the packages to install and the phase in which to do it.
Change-Id: Iae1ca4a112443860cd6ca8ce43378dd97c793338
A short time ago, devstack change the name of its lists of debian
packages from apts to debs. In order to properly pre-cache, we should
read these.
Change-Id: Id784e2876a2cbbede650b7e101ff90f7d74ec367
Allow all nodes to connect to each other on the fixed/private ips.
This change realated to:
https://review.openstack.org/#/c/114840/
Change-Id: I4b70f9184591d4563ce7b0a1b163b4055604f334
We see occasional name resolution failures on our tests slaves. Before
marking any of these nodes as ready in nodepool check that we can
resolve git.openstack.org and the local pypi mirror on that node.
This should help in preventing tests from failing with these errors by
catching them early and help in debugging by centralizing all occurences
of these failures in the nodepool logs.
Change-Id: I48ec87141687c1f6cd53eaada350fc5ce52cdef4
The current configure_mirror.sh expects no-one
want to use other than openstack.org pypi mirrors.
Having the configure_mirror.sh to use a costum,
NODEPOOL_PYPI_MIRROR instead of NODEPOOL_REGION when defined.
It allows averyone to use his local pypi mirror, with his
nodepool.
Change-Id: I6f0bd3372be09ccf423caacff83e615038c96ec5
Current centos7 builds on hpcloud are failing calling grub2-mkconfig;
from the logs:
+ sudo grub2-mkconfig -o /boot/grub2/grub.cfg
sudo: grub2-mkconfig: command not found
To be completely honest, I'm having a hard time explaining why this
works with rackspace images, but not hpcloud ones. I can not
replicate this in my own setup, even using the same hpcloud images.
Possibly it's something to do with the nodepool host environment when
it starts the script on the remote host? Without access to a failing
host I'm clutching at straws...
Change-Id: I9d9f759fac1a96cc785354ae5aa44398c908704d
During bashate -ing some files with .sh extension has been modified,
which are interpreted with /bin/sh, which takes a different syntax.
Fixing the issue:
- rename the files, so they are not picked up
- use syntax that's understood by sh
Change-Id: I617b24813df86687ca33745d0c940b0c659f2e14
The current HP cloud official centos6 images ("CentOS 6.3 Server
64-bit 20130116" - c1f9f404-39c2-4e83-b4bb-90d5ead16a36) by default
can't connect to mirrors.fedoraproject.org over https
---
[root@test ~]# urlgrabber -d 1,debug.txt https://mirrors.fedoraproject.org/metalink?repo=epel-6
* About to connect() to mirrors.fedoraproject.org port 443 (#0)
* Trying 140.211.169.197... * connected
* Connected to mirrors.fedoraproject.org (140.211.169.197) port 443 (#0)
* Initializing NSS with certpath: sql:/etc/pki/nssdb
* CAfile: /etc/pki/tls/certs/ca-bundle.crt
CApath: none
* NSS error -8092
* Closing connection #0
* SSL connect error
---
This requires an update of ca-certificates before it will work.
Change-Id: I9c5e595aa2b8e86188cb7db658ce4b630e6d0774
* nodepool/scripts/prepare_node.sh: The py3k-precise nodes use a
virtualenv which defaults to python3.3, and zuul has at least one
dependency which fails to install in this situation (python-daemon
tries to use a non-Py3K compliant print syntax in its
version_info.py module). By telling virtualenv to use python2
instead we sidestep this issue.
Change-Id: Idfa70dee91494da83380fa0c0a1f8427659f6ce2
We are renaming openstack-infra/config to
openstack-infra/system-config. This patch edits paths for
nodepool.
Change-Id: I75de3128e9d179b700df465726ed057439e5401e
Some vendor specific images have both -virutal and
-generic kernel images, so we need to remove both versions
of the troublesome -70 kernel so that we can install -69
and let openvswitch build.
Change-Id: Id6bdbefbf6fd3b783a189c1cf2ca0923ac5c43df
Fix:
E020: Function declaration not in format "^function name {$": 'function
log_extlinux() {'
...scripts/convert_node_to_xenserver.sh: L421
Change-Id: Ice9431275feba9e45f11e64d911976e643baf3c0
Add some filtering to 55-cache-devsatck-repos to handle centos7, which
actually gets setup with a DISTRO_NAME of rhel7 in devstack
Update the readme file to highlight the options for building other
images.
"DISTRO=centos7 ./tools/build-image.sh" creates a workable centos7
qcow image with this change and a d-i-b that includes
Ie24033468b78587ea87188ee1b843b26895798ff
Change-Id: Ida38cba0a303ab77432d5d354f1952e00ebfa00e
Also create tools/run-bashate.sh to wrap a find command to search for
files to run bashate on, and use it in tox.
Change-Id: I5ab07425f566a556f158a4b76b00aa35f9fb4385
Upstreaming several changes from XenServer CI:
- Use tty0 as console - thus the user can see what's happening during
installation over VNC console.
- Set the number of VCPUs to 6 for the DomU appliance
- Add prepare_node_xenserver.sh script.
- Use testing xva 1.1.4 which has trusty.
Change-Id: I56fc4e1af12f3f13fd1aa0b772bd8734b412d3d5
This repo was created from filter branching the openstack-infra/
config repo. This process brought a lot of cruft with it in the
form of directories that we no longer need. This patch removes
that cruft so we begin with a tidier repo.
Change-Id: Ibffad1b11c0c5f84eedfb0365369f60c4961a0f3