-
Notifications
You must be signed in to change notification settings - Fork 62
Conversation
Don't use the full labels (including the version) for the selector for the database.
BREAKING CHANGE: Maintaining the database integrity while doing Galera replication is currently not feasible in k8s as part of KubeCF. The general recommendation for doing HA database is to use an external database that supports this setup. Signed-off-by: Thulio Ferraz Assis <thulio.assis@suse.com>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please edit unsupported.yaml
to explicitly let users know that this is no longer supported.
/cc @fargozhu for release notes.
Co-authored-by: Mark Yen <3977982+mook-as@users.noreply.github.com>
Signed-off-by: Thulio Ferraz Assis <thulio.assis@suse.com>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Nitpicks around phrasing.
Is it safe to downgrade the database from HA to SA when upgrading kubecf? If not, then maybe the "unsupported" error should include some guidance to scale down the database before upgrading? (Assuming that is safe; or can that too still lead to data loss?) |
@jandubois Going from HA to SA should be safe if the active replica is the |
Co-authored-by: Mark Yen <3977982+mook-as@users.noreply.github.com>
@f0rmiga afaik it will always scale to the 0 ordinal |
Description
Removed the possibility of doing HA with the PXC database provided as part of KubeCF. It's currently not feasible to maintain a reliable HA setup as part of KubeCF.
Motivation and Context
When the active node doing writes is terminated in the middle of replication (e.g. going from SA to HA as part of an upgrade), the data is left broken.
How Has This Been Tested?
Locally. CI.
Types of changes
Checklist: