The block integration tests are a source of intermittent failures
in the CI. The PVCs were not being confirmed as removed before the
pools were deleted. Then the pool would not be removed since there
might still be a block image from the PVC that wasn't deleted yet.
Now the tests will ensure the PVCs and their block images are
removed before the pool is deleted.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
Creating an object user was always waiting at least 15 seconds before
attempting to create the user. With this change the operator will check
immediately if the object store is already initialized instead
of waiting for the initial 15 seconds delay to elapse.
Signed-off-by: travisn <tnielsen@redhat.com>
K8s 1.16 requires apps/v1 instead of the old extensions/v1beta1 for the
apiVersion of the deployment. The tests were failing when testing on
K8s 1.16 without this change
Signed-off-by: travisn <tnielsen@redhat.com>
Just as with the mon, mgr, mds, rgw, and rbd-mirror daemons, do not
generate a ceph.conf in an init container for directory-based OSDs.
Instead use the mon config database and the commandline to supply all
the needed arguments for running these OSDs.
Also generate a keyring secret to mount to OSD pods. Use this
secret for directory-based OSDs for now with the intention to
use this for other OSDs in the future.
Signed-off-by: Blaine Gardner <blaine.gardner@suse.com>
The tests must ensure that the cluster is cleaned up so that
other test suites will not be affected by cleanup of a previous
test suite. Removing the cluster finalizer is critical to this
cleanup.
Signed-off-by: travisn <tnielsen@redhat.com>
The filesystem test deletes the client test pod by deleting the entire yaml
and sometimes hangs the integration tests. This is the most common intermittent
error in the CI. Now the deletion will happen more directly with a pod delete
command. If it does still fail deletion at least the deletion should timeout
sooner.
Signed-off-by: travisn <tnielsen@redhat.com>
Also:
* `cluster.yaml`: spec.mgr.modules is not explicily nullable
* Removed `required: mon` in `ceph_manifests.go`
Signed-off-by: Sebastian Wagner <sebastian.wagner@suse.com>
There are six test suites and six k8s versions where we currently run
the tests. For efficiency we can restrict the testing to one suite
per k8s version. Bigger or riskier changes should still run the full
set of suites on all versions. To trigger the smaller test matrix, add
[test ceph min] to the PR description
Signed-off-by: travisn <tnielsen@redhat.com>
THe boolean helm settings were only being applied if their value was true.
If the desired value was false, the value would be skipped in the chart
instead of adding it with the value of false.
Signed-off-by: travisn <tnielsen@redhat.com>
Add support for MachineDisruptionBudget controller and MachineLabel controller required for fencing
in OpenShift. This will ensure that machines are only fenced and OSDs are only stopped
when Ceph is in a healthy state.
Co-authored-by: Ashish Ranjan <aranjan@redhat.com>
Signed-off-by: travisn <tnielsen@redhat.com>
We can now enable via the cluster CR any manager module with a
setting under the new mgr element
mgr:
modules:
- name: pg_autoscaler
enabled: true
Co-authored-by: Sébastien Han <seb@redhat.com>
Co-authored-by: travisn <tnielsen@redhat.com>
Signed-off-by: travisn <tnielsen@redhat.com>
- Renamed MDSCOUNT to ActiveMDS & updated Description
This was to remove the confusion about what the count
is actually about (Active or Active+Standby).
Signed-off-by: Umanga Chapagain <chapagainumanga@gmail.com>
After further consideration and working with the feature for a few
weeks, the `configOverride` functionality seems to be more complexity
than is worth it for the expected benefit. Users still have the most
preferred option of setting configs in Ceph's config database using
`ceph cofig set` or the dashboard, and the override ConfigMap still
exists for fallback scenarios. The config database can still be used
by Rook for setting necessary configs, but the user has no way of
telling Rook to do so.
Signed-off-by: Blaine Gardner <blaine.gardner@suse.com>
grpc metrics is made option in rook
this can be disabled/enabled with ROOK_CSI_ENABLE_GRPC_METRCS
variable in operator.yaml
this will help admin who want to check the grpc metrics
of the ceph-csi plugins
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
Allow users to specify overrides in the CephCluster CRD. The override
ConfigMap still exists for emergency situations and is mounted into
daemon pods directly instead of being merged into a Rook-created config
file.
Rook Ceph no longer uses a config file for managing daemons with the
exception of the OSDs which still generate a config in an init
container.
Signed-off-by: Blaine Gardner <blaine.gardner@suse.com>
The last image has disabled ephemeral repositories so it's now possible
for images older than 15 days to install packages without having an
error from non-existing repositories.
Closes: https://github.com/rook/rook/issues/3662
Signed-off-by: Sébastien Han <seb@redhat.com>
Deployment behaves better when a node gets disconnected
from the rest of the cluster - new provisioner leader
is elected in ~15 seconds, while it may take up to
5 minutes for StatefulSet to start a new replica.
if kube version is 1.13.x deploy provisioner as statefulset.
if kube version is higher than 1.14+ deploy provisioner
as deployment.
Refer: kubernetes-csi/external-provisioner@52d1fbc
Refer: ceph/ceph-csi#497
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
This is required if you are running multiple
csi drivers, if multiple node plugins are
listening on the same socket they request
may go to any node plugin. to avoid this we
need to have a unique socket per csi driver.
It is recommended to have socket inside the
driver name folder.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
Travis Nielsen
Sebastian Wagner
Sébastien Han
Blaine Gardner
RAJAT SINGH
Rohan CJ
Umanga Chapagain
Maksim Nabokikh
Madhu Rajanna
Anton Skriptsov
Noah Watkins
Sameer Kulkarni
jiffin