Add podman ceph-common and jq as part of preinstall dependency.
Add REMOTE_CEPH capabilities to CEPHADM deployment.
Removed set_min_client only if cinder is enabled, this should be set
in any case.
Get FSID from ceph.conf in /etc/ceph to avoid unnecessary override.
Part of an effort to test multinode deployments with cephadm.
Needed-By: I5162815b66d3f3e8cf8c1e246b61b0ea06c1a270
Change-Id: I84249ae268dfe00a112c67e5170b679acb318a25
This reverts commit 863a01b032.
Partial revert only for the pin to focal, leaves the broken other jobs
commented out.
Update paste-deploy workaround to be used always.
Add qemu-block-extra and podman deps to the debs list.
Running on the newer ceph and distro causes some quite different
performance characteristics that cause tests that used to pass to fail
more often. This includes some performance optimizations to help
reduce the memory footprint, as well as depends on changes to
tempest tests to improve the reliability of those tests by enabling
validation via SSH.
This also moves the cephadm job to be the voting/gating job as that
seems to be the clear consensus about "the future" of how we deploy
ceph for testing.
Depends-On: https://review.opendev.org/c/openstack/cinder-tempest-plugin/+/881764
Co-Authored-By: Dan Smith <dms@danplanet.com>
Change-Id: I899822fec863f43cd6c58b25cf4688c6a3ac1e9b
Add the option to deploy the Ceph cluster
with the cephadm tool.
Depends-On: I799521f008123b8e42b2021c1c11d374b834bec3
Co-Authored-By: Francesco Pantano <fpantano@redhat.com>
Change-Id: Id2a704b136b9e47b7b88ef586282cb5d0f754cf1
We were using Jewel/Luminous. Being a developer tool, we should
be sticking with newer versions.
This patch set also adds the following changes
- Refactor setup_packages_for_manila_on_<distro> functions
- Removes ceph_init call. Seems it lost relevance
- Removes APT repositories setup. We directly use shaman/chacra
packages directly
- Removes RUN_AS env var. It has been a while since Ceph
defaults to ceph user to run all the services. No longer needed.
- Removes sleep calls for services start checks. We use systemctl
checks.
- Adds more debugging logs
- Uses the search API from ceph shaman instead of hardwiring urls
- Adds the jq (JSON processor) dependency
Change-Id: I51c90e592070b99422e692d5e9e97083d93146e8
I5698e451861828a8b1240d046d1610d8d37ca5a2 introduced support for this
test in Stein so enable the compute feature from Stein onwards.
Blueprint: extend-in-use-rbd-volumes
Change-Id: I6b2575b12fdb9636f106b3ca28681e29ad3eb8ac
The dependent nova fix resolves the issue with rbd and
shelve/unshelve testing in tempest. Re-enabling the test
needs to be branch-specific since devstack-plugin-ceph
is branch-less but the fixes in nova will be branch
specific so we don't want to enable this on any nova
branch that does not have the fix since it would break
the job on those branches.
Related-Bug: #1653953
Depends-On: https://review.opendev.org/457886/
Co-Authored-By: Matt Riedemann <mriedem.os@gmail.com>
Change-Id: Ifbfc15033020c430a93574f2726b2279c9be130d
Similar to what happened some time ago for the stable/queens branch, we
are now seeing issues again on master after there was a new Paste
release being made a couple of weeks ago. So we need to perform
the same workaround of forcing pip to install PasteDeploy over
the distro version on the master branch.
Change-Id: I3e98a08e78182bdf5ec09b30e31b98435ec29456
Closes-Bug: 1803143
The workaround in order to fix the issue with Paste/PasteDeploy being
pulled in as system packages when installing ceph is only needed for the
Pike UCA. Assume that we only use that when devstack is running at the
stable/queens branch.
Change-Id: I0bbf04be015b777b3feb2c5e837b1f46f81cf3e9
This fixes a small logical error, we only want to set a libvirt secret
if nova is enabled, not both cinder and nova. This prevents a
deployment with cinder but not nova from being successful.
Change-Id: Ief83fe661cf1b2099af8131a7a206ffca3458dc7
The gate-tempest-dsvm-full-devstack-plugin-ceph-ubuntu-xenial job is
currently failing at a rate of 100% since we updated to use the Pike
UCA. The Ceph packages in the Pike UCA are pulling in python-paste and
python-pastedeploy. When devstack pip installs a newer version of
python-paste to satisfy upper-constraints, we end up with Paste
installed from pypi and python-pastedeploy installed from the Ubuntu
archive and Keystone fails to start with the error:
"ImportError: cannot import name deploy."
This adds a hack to force an update of PasteDeploy from pypi during the
devstack "stack install" phase to work around the problem and get the
job passing again.
Change-Id: Idf28de8faa56c2d387291c7d14ac6d4bf1c74c6a
This commit introduces the support of deploying a containerized ceph
cluster. It relies on the ceph/demo container image available as part of
the ceph-docker project (https://github.com/ceph/ceph-docker).
To enable this scenario, just set CEPH_CONTAINERIZED to True, all the
previous available options will work. So settings like
CEPH_LOOPBACK_DISK_SIZE continue to work.
Change-Id: Ie10155140448c04b88b9920381b54505f7359585
Signed-off-by: Sébastien Han <seb@redhat.com>
This patch enables the plugin to setup ceph radosgw and configures
it as a swift-compatible endpoint with keystone integration.
ENABLE_CEPH_RGW is the new variable introduced, which is False
by default, and can be set to True in localrc if radosgw needs
to be setup.
Fixed couple of other (related) things ...
1) Created rgw specific functions
2) Checking for radosgw & swift co-existence, erroring out early enough
NOTE: Tested on ubuntu trusty only
TODOs ...
1) Not tested on Fedora (F22), radosgw startup has some issues there.
Co-Authored-By: Deepak C Shetty <deepakcs@redhat.com>
Change-Id: I21db4168eb69d107599f6b6ab1668b02b764b2c6
The cleanup_ceph_embedded function nukes the ceph
conf directories, so when subsequently trying
to stop services they aren't stopped. Next stack.sh
run will fail because daemons are still running.
Fix by stopping daemons before running cleanup.
Change-Id: I51cd58eae83171a47aa898b65240b72ea0217af9
Closes-Bug: #1534600
Let the plugin setup and configure CephFS as the storage backend for
Manila. This is not done by default. Refer the README to do so.
Add 'pre_test_hook' and 'post_test_hook' scripts that would be needed
to configure Manila-Ceph CI jobs.
The following changes are also made:
* Install a development version of Ceph that is compatible with Manila's
CephFSNative driver.
* The development version is a version greater than Infernalis. And
from Infernalis onwards, the ceph daemons are run as user "ceph".
So allow the daemons to run as user "ceph" for such versions.
* Enhance get_ceph_version to check Ceph version even when the
ceph mon daemon is not up, by checking the CLI version.
Co-Authored-By: John Spray <john.spray@redhat.com>
Change-Id: I74314bfcc6b52d524bb84f2232a988f275b9afbf
Add configurables so that one need not have Ceph be the storage
backend for Cinder, Cinder-backup, Glance and Nova.
Let Ceph be the default storage backend for Cinder, Cinder backup,
Glance, and Nova. Allow the default setting to be overriden when the
global config, ENABLE_CEPH_{CINDER,C_BAK,GLANCE,NOVA} is set to
False in the local.conf/localrc file.
Closes Bug: #1521278
Change-Id: I3afe7c57d5d56f15fe533ad5b1dbe36ccd369e6b
Ashley Rodriguez