Problems with registry api "not starting" since update to CE 17.9 (jammy)

Hi

We updated via “apt” to 17.9.0-ce this afternoon and since then we’ve noticed 502 Gateway errors when trying to interact with the registry (was all working fine 24 hours prior)

I’ve trawled the logs but can’t see any errors (that wrt a daemon not starting) … just that the back end is no longer there …

`

2025/02/20 20:56:43 [error] 186709#0: *5778 connect() failed (111: Connection refused) while connecting to upstream, client: 2a01:2c0:a:540:9c4e:dcff:fe4a:6920, server: scm.morby.org, request: “GET /v2/custom-images/laravel-apache/manifests/latest HTTP/1.1”, upstream: “http://127.0.0.1:5000/v2/custom-images/laravel-apache/manifests/latest”, host: “scm.morby.org:5050

`

I’ve not changed configs, however in debugging I have tried a reconfigure, a restart, a reboot, and uncommenting and then commenting out various default config settings with “registry” in them … as well as setting things like debug_addr … but that doesn’t start a listener either.

Nothing on 5000 / 5005 / etc just the front end on 5050

Other than restoring from a backup from prior to the 17.8.x->17.9.0 update (messy and I’d rather not) … is there anything else I can try to resolve this?

Thanks in advance

Jon

I had meet similar issue.

I host a docker version gitlab with metadata database version container registry.
It works fine in 17.8.2, but fail after upgrade to 17.9.0.

There is some message in the registry log file (/var/log/gitlab/registry/current inside the container):

creating new registry instance: configuring application: there are pending database migrations, use the 'registry database migrate' CLI command to check and apply them

Whe try to run registry database migrate up, it will return the error:

configuration error: configuration path unspecified

The finial solution

  • run the command gitlab-ctl registry-database migrate up

The container registry works again.

3 Likes

Thanks, yes that did solve it

I looked everywhere for a hint as to what might be wrong and couldn’t find anything mentioned in logs or changelogs or anywhere :frowning:

1 Like

I also got the error:

creating new registry instance: configuring application: there are pending database migrations, use the ‘registry database migrate’ CLI command to check and apply them

After upgrading GitLab today to v18.0.0. However, despite the gitlab-ctl upgrade, gitlab-ctl reconfigure etc. It didn’t automatically apply the Registry database migrations ;( !?

Why would I need to run this manually, while the GitLab instance itself did do a migration already (and is up and running without issues), while the Registry is now giving 502 bad gateway since the registry is down.

Could somebody explain why sudo gitlab-ctl registry-database migrate up is not run automatically after a new GitLab APT (Omnibus) package update??

Or is that what you mean by: “No support for online migrations.”?