- By: Toni Schmidbauer ( Lastmod: 2021-10-16 )
We had the task of answering various questions related to upgrading Red Hat Quay 3.3 to 3.4 and to 3.5 with the help of the quay-operator.
Thankfully (sic!) everything changed in regards to the Quay operator between Quay 3.3 and Quay 3.4.
So this is a brain dump of the things to consider.
With Quay 3.4 the operator was completely reworked and it basically changed from opinionated to very opinionated. The upgrade works quite well but you have to be aware about the following points:
The name of the custom resource changed from
configHostname, used for providing the quay configuration UI, is no longer configurable
The password for the configuration UI is always regenerated after a configuration re-deployment
The volume size of the PostgreSQL PVC will change to 50G
In my test cluster I was using a 10G Ceph block device and the StorageClass did not support volume expansion. So my upgrade stopped at this point and I had to allow volume expansion in the storage class.
A horizontal pod autoscaler is also deploy during the upgrade. The default is to scale automatically to 20 pods, you might reconsider this…
With the Quay operator version 3.5 the operator is monitoring all namespaces for custom resources and needs to be installed in the openshift-operators namespace, this is how we upgraded Quay to 3.5 including the operator:
change the quay operator channel to 3.5
trigger an upgrade
now Quay gets upgraded to version 3.5
after the Quay upgrade you need to reinstall the operator:
deinstall the quay operator, Quay is not affected by this
reinstall the Quay operator (3.5) in all-namespaces
the re-installation of the operator triggers a quay deployment, all Quay pods are restarted!
You have to manually cleanup old
Red Hat is working on providing documentation on how to backup and restore Quay in various scenarios. There's an open task for this that provides more information https://issues.redhat.com/browse/PROJQUAY-2242.