ReportReconcileResult should never be given an object that is nil. It is
impossible to force compile-time checking for this because the object's
type is an interface. We can't force this, but we can rework the
function definition to handle more cases where the object may not be
returned fully-complete, and we can rework callers to return structs
(not pointers-to-structs) so it is less likely a caller will pass nil
to the function.
Signed-off-by: Blaine Gardner <blaine.gardner@redhat.com>
The canary pods are very short-lived pods and are not
running ceph in any form. The startup probe was mistakenly
enabled for the canary pod, which could cause the pod
to fail unexpectedly if the startup probe was triggered
after the delay timeout.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
msgr2 allows for encryption and/or compression across the wire.
Settings for enabling the encryption and compression are now
available in the cluster CR to that ceph will be automatically
configured with these settings when desired.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
Since Nautilus, mgr2 is already enabled by default. No longer
does the operator need to enable msgr2 after the mons are
configured or upgraded.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
If the csi driver was enabled at a later time, the mon endpoints
would not be populated. The populate the csi config, either a new
cluster reconcile or restarting the operator would be necessary.
It is simpler just to update the csi configmap with every reconcile
even if the csi driver is not enabled.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
If the pod namespace was not found by the csi helper to update
the mon endpoints configmap, the reconcile was failing. To allow
for the update to continue in unit tests and ignore the env var,
we only log the warning instead of failing the reconcile.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
When kube authentication is used, let's clarify the error message and
hopefully lead to a better understanding on where to look for.
It could be misleading to print "token auth" when using kube auth,
although internally a token is used (the JWT from the Service
Account).
Signed-off-by: Sébastien Han <seb@redhat.com>
When downscaling CephNFS clusters, be sure to remove secrets and
configmaps. Also be sure to only downscale for the currently-reconciling
CephNFS cluster.
Signed-off-by: Blaine Gardner <blaine.gardner@redhat.com>
If a specific disk is configured to have more than one OSD per device,
or if it has a metadata device specified, exclude it from raw mode on an
individual basis.
Signed-off-by: Blaine Gardner <blaine.gardner@redhat.com>
The controller for disruption budgets is verbose at cluster creation
when the cluster is not yet created, which generates a lot of noise
in the log that makes the cluster look unhealthy. This is an expected
condition, so we can change this logging to debug level.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
Currently, the admin/user can configure the ceph.conf
for daemon pods using https://rook.io/docs/rook/v1.7/
ceph-advanced-configuration.html#custom-cephconf-settings.
This the above custom ceph.conf is only for ceph pods.
the support to provide constom ceph.conf for cephcsi
is added in cephcsi PR 2476
This PR adds the support to create/update
ceph.conf for csi pods.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
The osd destroy command leaves the osd id in use instead
of fully purging the osd id. If the osd id is not also
removed, it gives the impression that there is still
something that needs to be cleaned up from the old osd.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
The purge job is intended to clean up all the resources
related to an OSD that is to be removed. An OSD with
a metadata, wal, and data PVC was only cleaning up the
data PVC. Now the metadata and wal PVCs will also be
cleaned up.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
If the filesystem is in a bad state such as never being
created successfully in the first place, the filesystem
removal will continuously fail and prevent the CR from
being removed. Thus, we treat the filesystem removal
as a best effort and log the issue instead of failing
the uninstall reconcile.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
Added unit test for probes changes and
did re-naming of liveness probes file
as it will contain multiple types of probes
Signed-off-by: parth-gr <paarora@redhat.com>
currently if multisite is not configured Rook sets values for
zone/zonegroup/realm as object store name which is only valid for
internal ceph object store. This may not exists for rgw running in
external cluster.
Signed-off-by: Jiffin Tony Thottan <thottanjiffin@gmail.com>
We don't need to validate the name of Ceph-CSI secret, simply because
the names can be different. The default example have default names but
since the secret can be created by the administrator we should not
assume anything. This check will fail if the secret names are different
for whatever reason. This is especially true for external mode. In
converged mode Rook creates the secrets on its own if needed to ease the
integration with the default StorageClass examples.
Signed-off-by: Sébastien Han <seb@redhat.com>
The external code has its own mechanism to look up the clusterInfo so
let's move this to the non-external condition to clarify the code.
Signed-off-by: Sébastien Han <seb@redhat.com>
This is fixing a potential race where the context is canceled and the
external cluster is trying to fetch the cluster info. If the context is
canceled then the loop will be stuck forever. Now we check for the
canceled context and return appropriately.
Signed-off-by: Sébastien Han <seb@redhat.com>
Instead of setting the default values in the Rook
we should use the defaults provided by the ceph
it self. This configuration is causing problem
for the rbd image creation, this is not allowing
the rbd images to have default image features
which comes with librbd.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
The OSD prepare job was succeeding even if the OSDs failed to be listed
by ceph-volume. These errors should be returned and fail the job, resulting
in a retry to retrieve the OSDs. Otherwise, the failure is masked and
the admin needs to dig through logs to find why the OSDs were not configured
on that node.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
The backend path was added with additional `transit` to it.
Also added PR test case to check transit engine
Signed-off-by: Jiffin Tony Thottan <thottanjiffin@gmail.com>
When a node is being skipped for creation of OSDs, log a message
that indicates whether it was from the node being unschedulable,
the node is not ready, or the node does not meet the placement
criteria specified in the cluster CR.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
We need to explicitly assign the ceph version within each scope since
one function returns a pointer and one is not.
Previously, the `desiredCephVersion` was recreated in the scope of the
`else` and thus a new variable `desiredCephVersion` was initialized (see
the `:=`).
Now, each scope creates and assign to clusterInfo a `desiredCephVersion`
value.
Signed-off-by: Sébastien Han <seb@redhat.com>
The object child controller should not compare versions if the object
store is external. The current version comparison is using the
cmdreporter to check the ceph version of the image in the CephCluster spec.
In external mode, this image is not set so the cmdreporter fails.
Also, this check is only valid for converged mode, so external mode
should be skipped.
Signed-off-by: Sébastien Han <seb@redhat.com>
In this case `oldMappingToGo` and `newMappingToGo` were not pointers so
there is nothing to dereference.
This happens when the monitors endpoints configmap is updated, but only
on the pool and file controllers. They are watching for changes on that
configmap so that they can reconcile and upate the bootstrap peers token
when replication is enabled.
Closes: https://github.com/rook/rook/issues/9714
Signed-off-by: Sébastien Han <seb@redhat.com>
If the KMS connection details for Vault have a VAULT_NAMESPACE we must
use it when building the vault client to auto-detect the KV internal
version.
Signed-off-by: Sébastien Han <seb@redhat.com>
This commit modifies ceph-csi controller to be able to cleanup
ceph-csi deployment,daemonset,services & csidriver objects
when no cephcluster exists.
It makes the following changes:
- redundant check for tp.DriverNamePrefix is removed.
- retry to start drivers within reconcile loop is removed,
controller will retry in case of error now.
- stopDrivers() will now return error in case of failure.
- predicate delete func will now respond to cephcluster deletion.
- CSI resources will be cleaned up when cephcluster does not exist.
Fixes: #9697
Signed-off-by: Rakshith R <rar@redhat.com>
The reconcile was skipping updating most pool properties for
object stores. The implementation of pools between the file,
object, and pool controllers had some duplicate code, so
this change also factors out the common code for better
reuse in a single place. Anytime a pool is created or updated,
it will now consistently update all the pool properties
that are expected to be modifiable.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
We can eliminate the OSD creation code for Ceph < v15.2.0 because Nautilus(v14.2.z)
is no longer supported. In other words, we can now expect that all newly-created
OSDs on PVC are raw mode.
Please note that the existing lvm mode OSDs on PVCs still be supported after this change.
Signed-off-by: Satoru Takeuchi <satoru.takeuchi@gmail.com>
Blaine Gardner
Travis Nielsen
Sébastien Han
parth-gr
Tom Hellier
Madhu Rajanna
Jiffin Tony Thottan
Satoru Takeuchi
Rakshith R
Mathieu Parent