Updated issue from 12.10.6 to 13.0.0 --> also can't downgrade anymore.

My gitlab instance was on the version 12.10.6 with the according redis and postgresqrl version as in the docker-compose of that date.

I tried to update to 13.0.0 together with the new redis and postgresql versions (again, the one from the docker-compose.yml), but the updated failed. For a strange reason, gitlab said I have to update to “11.11” first. So I tried to downgrade to 12.6 (backuped my docker-compose file) again, but then redis and/or postgresql don’t work anymore and restart immediately after docker-compose up.
Now my gitlab is not working and in a strange stage where I’m using the “newer” version of redis and postgresql, but still gitlab 12.10.6.
When accessing the gitlab instance via web, I get ERROR 500.
The logs of gitlab look fine. Postgresql gives me constantly the error:
STATEMENT: SELECT "personal_access_tokens".* FROM "personal_access_tokens" WHERE (revoked = false AND (expires_at >= NOW() OR expires_at IS NULL)) AND "personal_access_tokens"."token_digest" = $1 LIMIT $2
ERROR: column personal_access_tokens.token_digest does not exist at character 141

I wasn’t able to find any combination of the different instances that works anymore. Since my gitlab instance was quite big, I don’t want to recreate everything from scratch. Any suggestions?

Start by comparing the version-specific docker-compose.yml configurations used for GitLab 12.10.6 and 13.0.0. Inspect the PostgreSQL logs and the reported personal_access_tokens query, then trace the failed upgrade and downgrade state. Done means identifying a supported recovery path that restores the instance without recreating its data.