The osd prepare job detects the failure domain for the OSD by querying
the node topology labels. This topology is then assigned to the OSD
daemon in the CRUSH map. Previously, the affinity was required to be
set in the cluster CR, but it was very difficult to get right. Now the
operator will enforce the correct topology label on the OSD daemon
nodeAffinity by setting the label of the lowest topology in the hierarcy.
For example, if there are region, zone, and rack labels, the rack label
would be used to set the node affinity for the OSD daemon. If an
OSD prepare job is run in rack1, the corresponding OSD daemon will
have node affinity to rack1 to ensure the same topology. Previously,
the OSD could have ended up in rack2 unless the storageClassDeviceSet
placement was very carefully crafted.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
If the cluster was expanded we would iterate over the entire list of
OSDs again, create a new key and override in Vault... The Secret store
is not impacted since we have a handler for "already exist".
Signed-off-by: Sébastien Han <seb@redhat.com>
Add a new OSD provisioner status (reported by provisioner configmap)
that denotes that an OSD is preexisting and that OSD prepare does not
need to be run for the OSD. This only applies to OSDs on PVC currently
where existence of a deployment for the OSD indicates no further
provisioning needs to occur. New status is "preexisting".
Allow creating new OSDs on PVCs before updating existing ones by
allowing deferring OSDs during processing of OSD provisioning status
ConfigMaps. Deferral is identified by new "preexisting" status.
Add a unit test to ensure OSD on PVC provisioning occurs as expected
including deferring already-created OSDs to be updated after new PVCs
are created.
Signed-off-by: Blaine Gardner <blaine.gardner@redhat.com>
With 5f57b6fd3a, we introduced a
regression in the validation code.
If the OSD PV comes from another cluster, GetCephVolumeRawOSDs must
return an error.
The prepare job will crash loop but at least with a useful and
comprehensive message.
Signed-off-by: Sébastien Han <seb@redhat.com>
If the ceph cli outputs an error with "error calling conf_read_file"
this means that the operator has not written its ceph configuration
file. Thus ceph cli commands will fail, so we can just ignore that since
the operator will soon write this file in its initialization sequence.
Signed-off-by: Sébastien Han <seb@redhat.com>
When creating object store, `radosgw-admin realm get ..` command is stuck forever when required number of OSDs are not available. Because of this the uninstall of object store is also stuck. User has to manually remove the finalizer to delete the object store.This PR uses `ExecuteCommandWithTimeout` for running `radosgw-admin` command. Timeout during installation will be reconciled. Cleanup will be treated as best effort. Any errors during uninstalling of single site object store will only be logged.
Signed-off-by: Santosh Pillai <sapillai@redhat.com>
With Ceph Pacific, Rook can now deploy the cephfs-mirror daemon.
This initial commit covers the deployment of a single daemon only.
Multiple mirror daemons is currently untested.
Only a single mirror daemon is recommended.
The configuration of peers will come in a later PR since the mgr module
is still pending upstream: https://github.com/ceph/ceph/pull/39050
The same goes for integration tests, they will get added later once we
start testing on Pacific.
Closes: https://github.com/rook/rook/issues/7002
Signed-off-by: Sébastien Han <seb@redhat.com>
added waitTimeoutForHealthyOSD in the cluster crd that defines the time (in minutes) the operator would wait before an OSD can be stopped for
upgrade. This PR also removes the ok-to-continue logic for osds as its already handled by ok-to-stop. The default value is 10 minutes.
Signed-off-by: Santosh Pillai <sapillai@redhat.com>
The first patch to configure vault for ceph object store. If the `security.kms` configured in
`clusterSpec` CRD, RGW will be configured with vault kms settings to handle SSE request from s3 clients.
Signed-off-by: Jiffin Tony Thottan <thottanjiffin@gmail.com>
As of Pacific the balancer is now on by default in upmap mode.
In earlier versions, the balancer was included in the `always_on_modules` list, but needed to be
turned on explicitly using the ``ceph balancer on command.
Signed-off-by: Sébastien Han <seb@redhat.com>
14.2.15 lvm batch command prepare report changes
output format. This commit skips md check if ceph version
is greater than 14.2.13.
Signed-off-by: shenjiatong <yshxxsjt715@gmail.com>
This reverts commit a6ee5657ae.
The operator reconcile of the cephcluster was staying in an endless
loop in clusters that did not have OSDs configured on all nodes.
Now we revert this change for the patch release and will follow
up with a more complete fix in https://github.com/rook/rook/pull/6813.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
Test the code-flow when creating crush-rule via explicit
decompile/compile crushmap, when building two-steps crush rule. Relevant
for the case where 'PoolSpec.Replicated.ReplicasPerFailureDomain' is non
zeoro.
This test requires ceph's 'crushtool' to be installed on local machine.
Otherwise, it is ignored.
Signed-off-by: Shachar Sharon <ssharon@redhat.com>
When 'pool.Replicated.ReplicasPerFailureDomain' is non zero, the code
goes via route of adding explicit crush-rule from template, but ignores
the 'pool.DeviceClass' if it is set. Added option to template to have
(optional) 'class hdd|ssd'
Signed-off-by: Shachar Sharon <ssharon@redhat.com>
The deviceClass property was being ignored when creating the
non-pvc OSDs. Now the deviceClass will be specified as a property
for individual devices, all devices on a node, or all OSDs in the
cluster, depending on the level where the config is applied in the
cluster CR.
Co-authored-by: shenjiatong <yshxxsjt715@gmail.com>
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
The arbiter can only be configured with the stretch cluster if the
CRUSH map is balanced and there are two zones in the CRUSH map.
After the OSDs are configured, we wait for all the OSD pods to be
running and that the CRUSH map is balanced. If it takes more than
two minutes, we fail the reconcile and try again. This is only done
the first time the stretch cluster is configured. In future reconciles
we first check if the stretch cluster is already enabled before
enabling it again.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
The code was assuming that devices were passed by the user as
"/dev/sda", this is bad! We all know people should be using paths like
/dev/disk/by-id so we must support them.
Closes: https://github.com/rook/rook/issues/6685
Signed-off-by: Sébastien Han <seb@redhat.com>
Now, we can schedule snapshots on pools from the CephBlockPool CR when
the pool is mirrored.
It can be enabled like this:
```
mirroring:
enabled: true
mode: pool
snapshotSchedules:
- interval: 24h # daily snapshots
startTime: 14:00:00-05:00
```
Multiple schedules are supported since snapshotSchedules is a list.
Signed-off-by: Sébastien Han <seb@redhat.com>
When new pools are added to an already existing filesystem, we would
create them but not actually associate them to the filesystem. Ensure
that we do this.
Also, while we're here, we refactor the CreateFilesystem test: before,
we were not properly testing creating the filesystem from scratch
because we would return valid JSON for *every* ceph command. Change this
by making the first run return an error so that we get some better
coverage.
After this, we also add a check for the correctness of adding new pools
to the CreateFilesystem unittest.
Fixes#5876
Signed-off-by: Lalit Maganti <lalitm@google.com>
Found by running the following command:
codespell -S .git,*.png,*.jpg -L \
aks,keyserver,atleast,dne,ser,ist,files\',ba,dum,iam,te -f -H
Signed-off-by: Mateusz Gozdek <mgozdekof@gmail.com>
Pools in stretch clusters will all use the same crush rule that
is generated by the operator when configuring the cluster for
stretch mode, with replica 4 and two replicas per failure domain.
Pools cannot create new crush rules, so we require that replica be
4 when the pools is created, EC is not allowed, and no new
crush rule will be created.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
In clusters where only two datacenters (or similar failure domains)
are available, a different mon and osd approach is needed to deal
with the network partitions or some other reason for one of the failure
domains going down. The Ceph stretched cluster makes the mons aware
of the failure domains by configuring one as the arbiter in a third
zone, while keeping two replicas of the data in each of the data
zoens.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
The osd purge job was not removing the osd prepare job or the
osd pvc due to an invalid label query for the pvc. Now the
prepare job and pvc will be removed as expected from the job.
Signed-off-by: Travis Nielsen <tnielsen@redhat.com>
Golang maps order is not always guaranteed to be identical, which makes
the table driven test to fail. Because of the nature of the maps, it is
hard to predict the exact position of an element in the map. This
makes the table driven test flappy.
Reworked the unit test using a contains statement which achieves the
same validation.
Closes: https://github.com/rook/rook/issues/6522
Signed-off-by: Sébastien Han <seb@redhat.com>
When the Ceph cluster runs on PVC and the OSDs are encrypted we can
store LUKS's Key Encryption Key inside a Key Management System. Today,
Rook only supports HashiCorp Vault: https://www.vaultproject.io/
The CephCluster has now a new "security" field which will plug onto the
KMS. Here is an example:
security:
kms:
tokenSecretName: <name of the secret containing a Vault token, used
to authenticate>
connectionDetails: < a map of strings containing connection
information>
Refer to the ceph-cluster-crd documentation to lear more.
Closes: https://github.com/rook/rook/issues/6105
Signed-off-by: Sébastien Han <seb@redhat.com>
The CRUSH map does not allow duplicate values (nor keys) for the
labels. That means that if the currently hardcoded label value
`default` occurs anywhere else in the tree (for example, because
some of the topology labels provided by the cloud provider use it),
the cluster cannot sucessfully spawn.
By giving the user control over the label value used for the root
CRUSH map label, it is possible to adapt the cluster to this
situation.
To be able to actually use the `default` value, the root=default
node which is created by default by ceph itself has to be removed;
since that node is accompanied by a replicated crush rule, we have
to remove that rule, too.
The integration tests are modified to sometimes use the custom
crushRoot (based on an arbitrary criterium) to get coverage across
the various suites. Separate tests could be added, but were not
deemed necessary at this point.
Fixes#4993.
Signed-off-by: Jonas Schäfer <jonas.schaefer@cloudandheat.com>
golangci-lint linter gosec showing more errors
than gosec gh. This commit resolve
new errors. And, removing
nosec comments from autogenerated files.
Signed-off-by: subhamkrai <srai@redhat.com>
Sébastien Han
Travis Nielsen
Frank Ritchie
crombus
Blaine Gardner
Santosh Pillai
Jiffin Tony Thottan
shenjiatong
Shachar Sharon
Stefan Haas
Arun Kumar Mohan
Lalit Maganti
Mateusz Gozdek
Jonas Schäfer
subhamkrai
Daichi Sakaue