This commit is a large refactor on how the operator starts, stops and
how it starts various sub-components such as the ceph-csi driver. It
also refines the way we cancel orchestrations. We don't use breakpoints
anymore but send our self a SIGUP to reload our controller runtime
manager.
The reload will happen under different circonstances like:
* a new adminission controller secret is created/deleted/changed
* a CephCluster CR is edited
As mentioned earlier, the csi driver now has its own controller, just
like flex. It reacts to change in the operator config map for particular
ROOK_CSI_ fields.
A second new controller for the operator's general config has been
created, it manages:
* the logging level
* the ceph CLI command timeout
* the discovery daemon
The operator reacts much more rapidly to cancellation events by stopping
the manager's context and reloading it.
Signed-off-by: Sébastien Han <seb@redhat.com>
Let's force v16.2.5 since the CI is broken with 16.2.6. This gives us
time to continue to merge work and work on fixing deployments with
16.2.6 in parallel.
Signed-off-by: Sébastien Han <seb@redhat.com>
The release version needs to be the same in the example/test manifests
as it is in the local build image. The github actions are different in
this regard than the Jenkins builds were. The Jenkins builds always
locally used the master tag instead of a release-specific tag.
Now that the github actions use the release-specific tag, the test
framework no longer should be using the master tag in release branches.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
Our admission webhooks will now run as part of the Operator container
and not an additional deployment. This has the advantage of consuming
fewer resources in the cluster and not having to manage affinities and
tolerations. This only drawback is that the Secret containing the
certificates is not mounted anymore and the content needs to be written
inside the Operator. This is not practical since we also need to watch
for the Secret content to change. Meaning that the certificates have
been renewed and the webhook server needs to use them.
A new approach is on its way to hopefully simplify this last issue and
implement a watcher for the Secret.
In the meantime, users need to use the cert-manager or renew
certificates manually. Additionally, they must update the
ValidatingWebhookConfiguration object with the new CA bundle.
Signed-off-by: Sébastien Han <seb@redhat.com>
Recently lib-bucket-provisioner add support for update() API.
Include that on the obc implementation since it can be used to
update quota for OBC.
Fixes: #7146
Signed-off-by: Jiffin Tony Thottan <thottanjiffin@gmail.com>
This commit adds an ingress resource to the rook-ceph-cluster helm chart, allowing
ingress to the ceph-dashboard service. It also adds the ability to define the
various storage types that you can run on ceph inside kubernetes.
Closes https://github.com/rook/rook/issues/8384
Signed-off-by: Tom Hellier <me@tomhellier.com>
Enable again the mgr test:
- Now is more reliable and robust the start of the test.
- Minor fixes to adapt the <service ls> to the new name of the crash daemon
deployed by rook
- The creation of OSDs is disabled in the orchestrator, so i have removed
this test until we will have the functionality ready again in the orchestrator
part
My plan is to provide in the orchestrator two different ways to create OSDs:
- Creation of OSD using specific devices if discovery daemon is running
- Creation of OSds using PVS (if we have LSO/other LS operator running)
Signed-off-by: Juan Miguel Olmo Martínez <jolmomar@redhat.com>
Recently, the builds of `ceph/ceph` image moved to quay.io, see
https://github.com/ceph/ceph-build/pull/1883 for more details.
Current images will remain but new builds will happen on quay.io only.
This means that tags such as `v14.2`, `v15.2`,`v16.2` will need to
switch to quay.io to get updates.
Signed-off-by: Sébastien Han <seb@redhat.com>
With v1.7 approaching, the upgrade integration test will now test from
v1.6.x to the latest master, which will effectively become the v1.7
release soon.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
The CRDs v1 requires the full schema for all settings, so we now
generate the CRDs for nfs for full fidelity of all settings.
Co-authored-by: Nicolaj Græsholt <figaw@hotmail.com>
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
The CRDs v1 requires the full schema for all settings, so we now
generate the CRDs for cassandra for full fidelity of all the
settings.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
The ceph provider had been refactored to read the crds
and operator manifests from a file instead of copying the
manifests into the test code. Now the helpers are refactored
to allow the other providers to also reduce the manifest
duplication in tests.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
When the CephCluster is configured with Multus and multiple networks are
used to deploy Ceph some commands are failing to be executed from the
Operator. These commands, in particular, `radosgw-admin` ones need access
to the "ceph public network" to talk to OSDs. Unfortunately, the
Rook-Ceph Operator does not have the network annotations and thus
doesn't have the networks available and cannot reach OSDs. So the commands end
up hanging and eventually time out.
Applying the annotations to the Operator pod is possible but will result
in restarting the operator too and this should be avoided at all costs.
Also, applying the annotations beforehand is not possible since the
Multus declaration is in the CephCluster specification. So we would have
no idea what to do.
So the current approach runs a new sidecar container in the mgr pod to
act as a proxy for "some" ceph commands, only the `radosgw-admin` ones
for multi-site setup. This is a small container with admin access
running idle waiting for commands to be executed. In a sense, it is
similar to the toolbox but we didn't want to clearly expose it, so
running as a sidecar is quite nice.
Proxying command is obviously not always recommended since we add an
extra hop in the network path. Now each request has to go from the
operator pod to the API server to the remote pod to Ceph. Previously,
the command only goes from the operator to Ceph.
It's worth noting that external mode is not impacted since no rgw pod
is configured. This scenario is flexible and allows us to scale
pretty well since any CephCluster with Multus will see its mgr sidecar
deployed and can then talk to Ceph. We are not limited.
Signed-off-by: Sébastien Han <seb@redhat.com>
The upgrade test needs to also update the toolbox to ensure that it will not
be denied access to the cluster when the insecure connections are disabled
by the operator.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
Adds rook-ceph-cluster chart to chart build, add tests.
Pulled out some common functionality between the Helm and non-Helm installers
Signed-off-by: Henry Zhang <me@henry.dev>
`CephObjectStore->gateway->type` is not used. Rook has only supported s3-like
interface and hasn't had no code which handles `type` field.
In addition, this field was removed from CRD in the following commit.
ceph: auto-gen crds
31db03fece
Signed-off-by: Satoru Takeuchi <satoru.takeuchi@gmail.com>
The multicluster test was never removing the finalizer of the external
cluster since the core cluster was being removed first. Now we remove
the external cluster first to ensure the finalizer will be removed
properly instead of forcefully by the test.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
This commit updates external-snapshotter version to
v4.0.0 which supports snapshots v1.
Rook now defaults to enabling RBD and CephFS snapshotter
for K8s >= v1.17 and disabling it for K8s <= v1.16.
Supporting changes in documents and examples yaml files
are made.
Signed-off-by: Rakshith R <rar@redhat.com>
Because of the recent CRD changes made, the mon count had a minimum of
1, making the configuration of the external cluster impossible. The
operator would fail to add the finalizer:
```
2021-04-29 16:57:38.429451 E | ceph-cluster-controller: failed to reconcile. failed to add finalizer: failed to add finalizer "cephcluster.ceph.rook.io" on "test-external": CephCluster.ceph.rook.io "test-external" is invalid: spec.mon.count: Invalid value: 0: spec.mon.count in body should be greater than or equal to 1
```
We now fixed the CRD as well as adding the CR status.
Signed-off-by: Sébastien Han <seb@redhat.com>
We have seen new cases where retrying to lock the device 3 times is not
enough. It's the same race we had experienced where Ceph tries to
acquire a lock on the device but systemd-udevd does the same too.
Retrying 20 times every 0.1sec seems to mitigate that issue in the CI.
Signed-off-by: Sébastien Han <seb@redhat.com>
Since Rook 1.6 is using raw mode for simple OSD scenarios we can use v14
again without having issue with ceph-volume.
We keep the upgraade test with v14.2.12 since Rook 1.5 does not have the
raw code to deploy OSDs.
Closes: https://github.com/rook/rook/issues/7669
Signed-off-by: Sébastien Han <seb@redhat.com>
Rook's testing utilities include a function for loading the default
kubeconfig for a cluster. This function requires maintainance effort and
also doesn't support authentication methods like Basic Authentication.
Instead of adding support for it, drop the config loader and use the one
provided by the controller-runtime library.
Signed-off-by: Yannis Zarkadas <yanniszark@arrikto.com>
The integration tests pick up the manifests directly from the examples
folder, including with the release tag. In the release branch the Jenkins
build still uses the master build when building locally before tagging
it with the release tag. So the tests need to use the master tag.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
(cherry picked from commit 456e2f3a79)
With the Ceph Pacific release coming this week we add support
in Rook for Pacific with the Rook v1.6 release coming soon.
The integration tests will now run across nautilus, octopus,
and pacific to cover all supported Ceph versions. The default
examples still specify Octopus until there is more bake time
for Pacific.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
If a CI test fails we don't want to cleanup anything and leave the
cluster in the state it is. This will help debugging the CI.
Signed-off-by: Sébastien Han <seb@redhat.com>
The integration tests run independently in the github actions so there is
no need to cleanup from every test. The cleanup is still needed in the
Jenkins tests where all the suites run serially.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
The integration tests have long been painful to maintain with
settings in various places and copied to multiple types,
inconsistent variable names, and otherwise difficult to maintain
code. Now the settings for a test suite are all in one place and
they remain in the same settings type throughout the test.
The multi-cluster suite is also refactored to use the same install
and uninstall helpers as the other suites.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
The rbac required for multus was present in role which makes
only the network-attachment-definitions in the rook cluster namespace
accessible to the operator. Moved it to clusterrole for cluster wide
access to NAD.
Signed-off-by: rohan47 <rohgupta@redhat.com>
It's better to validate ownerReferences when setting them. In addition, we should use
controllerrutil.Set{Controller,Owner}Reference, that have such validation, as possible.
Signed-off-by: Satoru Takeuchi <satoru.takeuchi@gmail.com>
Sébastien Han
Jiffin Tony Thottan
Blaine Gardner
Travis Nielsen
parth-gr
Tom Hellier
Juan Miguel Olmo Martínez
Andy Bursavich
subhamkrai
Henry Zhang
Satoru Takeuchi
Rakshith R
bipuladh
Yannis Zarkadas
rohan47