Yes, we have 5 times the same repo, because of security or something. You are supposed to merge from 0 to 4. If you merge 2 commits into 3 and another one into 4, we are going to have a little problem.
The setup could make sense, if there’s a separate repo that’s pull only. Our ops guys pull our ArgoCD repo, so we can do the actually work, but they control what’s deployed on production.
Yes, but Gitlab doesn’t allow for easy access rules.
Basically, OPS wants full control of the repo, since they are the ones being blamed if something goes wrong. There’s no way to enforce, that only a certain set of users can make changes to a branch - all such restrictions can be circumvented rather easily.
So the solution is a shadow copy of the repo that only gets updated on release and Argo only deploys a specific tag (i.e. release).
We’re not talking about just some enterprise microservice, but stuff in the public administration/government sphere. The tradeoffs are a bit different there.
I didn’t know that GitLab doesn’t allow that! We use BitBucket and there it’s extremely easy to put branch restrictions so that only certain Usergroups are allowed to merge into the release-branches
Bitbucket also doesn’t enforce these rules properly. You can simply change the rules, merge, then change back.
The only way around that is to restrict every developer account into oblivion and only have an ops guy as repo admin, but I think most ops teams have better things to do.
That very much depends on your workflow and team setup. Repo admin for me means “can alter who and how branches can be merged”. That’s usually a job for lead devs.
Yes, we have 5 times the same repo, because of security or something. You are supposed to merge from 0 to 4. If you merge 2 commits into 3 and another one into 4, we are going to have a little problem.
That’s the dumbest solution ever.
I know that you are probably not the one who came up with this shit but just… why? Why would you do that?
Don’t you know that git has so called branches?
The setup could make sense, if there’s a separate repo that’s pull only. Our ops guys pull our ArgoCD repo, so we can do the actually work, but they control what’s deployed on production.
However, this seems not to be the case here.
But… also in ArgoCD you just set up which branch you want to look at
Yes, but Gitlab doesn’t allow for easy access rules.
Basically, OPS wants full control of the repo, since they are the ones being blamed if something goes wrong. There’s no way to enforce, that only a certain set of users can make changes to a branch - all such restrictions can be circumvented rather easily. So the solution is a shadow copy of the repo that only gets updated on release and Argo only deploys a specific tag (i.e. release).
We’re not talking about just some enterprise microservice, but stuff in the public administration/government sphere. The tradeoffs are a bit different there.
I didn’t know that GitLab doesn’t allow that! We use BitBucket and there it’s extremely easy to put branch restrictions so that only certain Usergroups are allowed to merge into the release-branches
Bitbucket also doesn’t enforce these rules properly. You can simply change the rules, merge, then change back.
The only way around that is to restrict every developer account into oblivion and only have an ops guy as repo admin, but I think most ops teams have better things to do.
Why should a developer be a Repo admin? Thats DevOps territory
That very much depends on your workflow and team setup. Repo admin for me means “can alter who and how branches can be merged”. That’s usually a job for lead devs.
Aah, DevOps as a separate role… Now that’s a dream I can get behind…