You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[enterprise-4.9] Issue in file registry/configuring_registry_storage/configuring-registry-storage-osp.adoc #43053
Yes @maxwelldb, you're correct - since the new backend is RWO, we need to ensure the image registry runs a single replica and that the rollout strategy is Recreate. On environments where Object Storage (Swift) is not available, the cluster-image-registry-operator sets this up automatically for you, however this would be an issue if someone follows the above procedure to migrate from a Swift backend (RWX) to Cinder (RWO).
We should add a note about the limitation when using RWO storage and change the example oc patch command to also set the replica count and rollout strategy accordingly (not tested):
Mark the issue as fresh by commenting /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.
Exclude this issue from closing by commenting /lifecycle frozen.
If this issue is safe to close now please do so with /close.
Mark the issue as fresh by commenting /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.
Exclude this issue from closing by commenting /lifecycle frozen.
If this issue is safe to close now please do so with /close.
Mark the issue as fresh by commenting /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.
Exclude this issue from closing by commenting /lifecycle frozen.
If this issue is safe to close now please do so with /close.
Mark the issue as fresh by commenting /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.
Exclude this issue from closing by commenting /lifecycle frozen.
If this issue is safe to close now please do so with /close.
Mark the issue as fresh by commenting /remove-lifecycle rotten.
Rotten issues close after an additional 30d of inactivity.
Exclude this issue from closing by commenting /lifecycle frozen.
If this issue is safe to close now please do so with /close.
Mark the issue as fresh by commenting /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.
Exclude this issue from closing by commenting /lifecycle frozen.
If this issue is safe to close now please do so with /close.
Mark the issue as fresh by commenting /remove-lifecycle rotten.
Rotten issues close after an additional 30d of inactivity.
Exclude this issue from closing by commenting /lifecycle frozen.
If this issue is safe to close now please do so with /close.
ReadWriteOnce means the storage can only be mounted on a single node but the node spreading priority will attempt to place the registry image on different nodes. The issue with corruption seems to be limited to specific NFS implementations using Filesystem type but, if this issue doesn't exist on Cinder or CephFS then ReadWriteMany for the Filesystem type is an alternative to constraining the registry scale for multi-node access.
Activity
maxwelldb commentedon Mar 11, 2022
@mandre @pierreprinetti Would you be able to confirm this? I believe the reference URL is https://docs.openshift.com/container-platform/4.9/registry/configuring_registry_storage/configuring-registry-storage-osp.html.
mandre commentedon Mar 21, 2022
Yes @maxwelldb, you're correct - since the new backend is RWO, we need to ensure the image registry runs a single replica and that the rollout strategy is Recreate. On environments where Object Storage (Swift) is not available, the cluster-image-registry-operator sets this up automatically for you, however this would be an issue if someone follows the above procedure to migrate from a Swift backend (RWX) to Cinder (RWO).
We should add a note about the limitation when using RWO storage and change the example
oc patchcommand to also set the replica count and rollout strategy accordingly (not tested):maxwelldb commentedon Mar 21, 2022
@mandre Thanks!
openshift-bot commentedon Jun 19, 2022
Issues go stale after 90d of inactivity.
Mark the issue as fresh by commenting
/remove-lifecycle stale.Stale issues rot after an additional 30d of inactivity and eventually close.
Exclude this issue from closing by commenting
/lifecycle frozen.If this issue is safe to close now please do so with
/close./lifecycle stale
maxwelldb commentedon Jun 19, 2022
/remove-lifecycle stale
openshift-bot commentedon Sep 18, 2022
Issues go stale after 90d of inactivity.
Mark the issue as fresh by commenting
/remove-lifecycle stale.Stale issues rot after an additional 30d of inactivity and eventually close.
Exclude this issue from closing by commenting
/lifecycle frozen.If this issue is safe to close now please do so with
/close./lifecycle stale
maxwelldb commentedon Sep 19, 2022
/remove-lifecycle stale
openshift-bot commentedon Dec 19, 2022
Issues go stale after 90d of inactivity.
Mark the issue as fresh by commenting
/remove-lifecycle stale.Stale issues rot after an additional 30d of inactivity and eventually close.
Exclude this issue from closing by commenting
/lifecycle frozen.If this issue is safe to close now please do so with
/close./lifecycle stale
3 remaining items
openshift-bot commentedon Mar 20, 2023
Issues go stale after 90d of inactivity.
Mark the issue as fresh by commenting
/remove-lifecycle stale.Stale issues rot after an additional 30d of inactivity and eventually close.
Exclude this issue from closing by commenting
/lifecycle frozen.If this issue is safe to close now please do so with
/close./lifecycle stale
openshift-bot commentedon Apr 19, 2023
Stale issues rot after 30d of inactivity.
Mark the issue as fresh by commenting
/remove-lifecycle rotten.Rotten issues close after an additional 30d of inactivity.
Exclude this issue from closing by commenting
/lifecycle frozen.If this issue is safe to close now please do so with
/close./lifecycle rotten
/remove-lifecycle stale
openshift-bot commentedon Jul 19, 2023
Issues go stale after 90d of inactivity.
Mark the issue as fresh by commenting
/remove-lifecycle stale.Stale issues rot after an additional 30d of inactivity and eventually close.
Exclude this issue from closing by commenting
/lifecycle frozen.If this issue is safe to close now please do so with
/close./lifecycle stale
openshift-bot commentedon Aug 18, 2023
Stale issues rot after 30d of inactivity.
Mark the issue as fresh by commenting
/remove-lifecycle rotten.Rotten issues close after an additional 30d of inactivity.
Exclude this issue from closing by commenting
/lifecycle frozen.If this issue is safe to close now please do so with
/close./lifecycle rotten
/remove-lifecycle stale
bbeaudoin commentedon Aug 18, 2023
/remove-lifecycle stale
bbeaudoin commentedon Aug 18, 2023
The issue with the OSP documentation being incomplete still potentially exists with 4.13: https://docs.openshift.com/container-platform/4.13/registry/configuring_registry_storage/configuring-registry-storage-osp.html
ReadWriteOnce means the storage can only be mounted on a single node but the node spreading priority will attempt to place the registry image on different nodes. The issue with corruption seems to be limited to specific NFS implementations using Filesystem type but, if this issue doesn't exist on Cinder or CephFS then ReadWriteMany for the Filesystem type is an alternative to constraining the registry scale for multi-node access.
The documentation is correct for bare metal given RWO storage but is also required to be RWO due to block requires special handling at https://docs.openshift.com/container-platform/4.13/registry/configuring_registry_storage/configuring-registry-storage-baremetal.html#installation-registry-storage-block-recreate-rollout-bare-metal_configuring-registry-storage-baremetal