How can I upgrade on prem docker?

[asm-ch]

I am running 1.9.3 and when I upgrade the container to e.g. 1.10.3 the cluster is gone. The 3 nodes are still present.
I try to enter the web page "Clusters" but it doesn't seem to work, nothing is displayed there. Neither "Machines" or "Installation media".
Any hints?

[MLuc24]

Good news first: a running container with an empty UI almost certainly means your data is intact and Omni is reading somewhere else, rather than anything having been destroyed.

You can rule out migrations straight away. Omni runs them during state setup and propagates any failure upward:

if _, err := migration.NewManager(resourceState, logger.With(logging.Component("migration"))).Run(ctx); err != nil {
    return err
}

so a broken or partial migration aborts startup rather than leaving you with a healthy container and no clusters. There's also a mechanism where dropped migrations refuse to run and tell you to upgrade to an intermediate version first — but that one's pinned at v1.1.0, so it can't apply to a jump from 1.9.3. Both of those would be loud, and neither matches what you're seeing.

What does match is Omni opening a store that's empty. An empty state isn't an error: migrations have nothing to do, startup succeeds, and every page renders zero rows. Clusters, Machines and Installation media all being blank at once is the tell — that's not three features failing, it's one datastore with nothing in it.

Worth knowing there's more than one place it could be looking. Alongside the embedded etcd path (EmbeddedDBPath), newer versions have a SQLite backend configured through storage.sqlite.path, or --sqlite-storage-path on the command line, and it's a required setting when that type is selected. So the thing to compare is which storage type and path your 1.10.3 container is running with versus what 1.9.3 wrote to, and whether the volume holding the old data is still mounted at the path the new one expects.

The container log at startup should say which backend it opened and where — that'll settle it in one line. If it names a path that isn't your old volume, nothing is lost and it's a mounting/config fix rather than a recovery job.

Whatever you do, take a copy of the old data directory before experimenting further, while it's still untouched.

[asm-ch]

Sorry, did not understand much.
Since this is test and I wanted to begin from scratch I deleted the sqlite folder content.
Redeployed container v1.10.3 and now it's working again.