-
Bug
-
Resolution: Done
-
Blocker
-
fuse-7.8-GA
During the upgrade process, the db-backup pod get stuck in ImagePullBackOf due to auths to the `registry.connect.redhat.com` repo:
Events: Type Reason Age From Message ---- ------ ---- ---- ------- Normal Scheduled 11s default-scheduler Successfully assigned mkralik/db-backup-1607343267-4gr6p to node1.fo-311-c.dos.fuse-qe.eng.rdu2.redhat.com Normal Pulling 8s kubelet pulling image "registry.connect.redhat.com/crunchydata/crunchy-pgdump:ubi7-12.4-4.5.0" Warning Failed 6s kubelet Failed to pull image "registry.connect.redhat.com/crunchydata/crunchy-pgdump:ubi7-12.4-4.5.0": rpc error: code = Unknown desc = Get https://registry.connect.redhat.com/v2/crunchydata/crunchy-pgdump/manifests/ubi7-12.4-4.5.0: unauthorized: Please login to the Red Hat Registry using your Customer Portal credentials. Further instructions can be found here: https://access.redhat.com/RegistryAuthentication Warning Failed 6s kubelet Error: ErrImagePull
The install process creates a syndesis-pull-secret with credentials to `registry.redhat.io`
e.g.:
{ "auths":{ "registry.redhat.io":{ "username":"<USERNAME>", "password":"<PASSWORD>", "auth":"<AUTH>" } } }
but the db-upgrade pod uses images from `registry.connect.redhat.com` . In that case, the install process should also create credentials for `registry.connect.redhat.com` as well. (It can use the same values as for `registry.redhat.io`).
{ "auths":{ "registry.redhat.io":{ "username":"<USERNAME>", "password":"<PASSWORD>", "auth":"<AUTH>" }, "registry.connect.redhat.com":{ "username":"<USERNAME>", "password":"<PASSWORD>", "auth":"<AUTH>" } } }
When I update .dockerconfigjson in syndesis-pull-secret manually (decode it, add registry.connect.redhat.com repo to the json with the same auths as registry.redhat.io repo), the image is pull sucesfully.
Strange is that I don't have this problem on OCP 4.x even though my syndesis-pull-secret contains credentials only for `registry.redhat.io` and doesn't contain credentials for `registry.connect.redhat.com`
{ "auths": { "quay.io": { "auth": "<AUTH>" }, "registry.redhat.io": { "auth": "<AUTH>" } } }
Events: Type Reason Age From Message ---- ------ ---- ---- ------- Normal Scheduled 89s default-scheduler Successfully assigned mkraliktestinstall12/db-backup-1607351006-tdhjs to crc-tnpk6-master-0 Normal AddedInterface 87s multus Add eth0 [10.116.0.82/23] Normal Pulling 87s kubelet Pulling image "registry.connect.redhat.com/crunchydata/crunchy-pgdump:ubi7-12.4-4.5.0" Normal Pulled 56s kubelet Successfully pulled image "registry.connect.redhat.com/crunchydata/crunchy-pgdump:ubi7-12.4-4.5.0" in 31.024212373s Normal Created 55s kubelet Created container backup-db-compiler
I see two possibilities how to fix it.
- The `create_secret_if_not_present` function needs to be updated to contains also auths to `registry.connect.redhat.com` *
- Or, we will use images from registry.redhat.io in case it is possible.
- - Another problem here is that the customers, which use 7.6, has the syndesis-pull-secret already installed without auths to `registry.connect.redhat.com` so they need to delete it before the upgrade process. The upgrade process creates a new one. This can be documented as known limitation during the upgrade 7.6->7.7 on OCP 3.11
- is caused by
-
ENTESB-15364 Use productized images for backup
- Done