Hello.
We are stuck in endless discussions trying to redesign our pipeline. We trust in the power of community help, so I am writing this message.
Some background data.
We use bitbucket cloud. A pipeline we have does both CI and CD. All apps are deployed to kubernetes. At this point in time we have to keep using bitbucket pipelines, no separation is possible. Our repos are combination of the source code and helm chart. Let's called it monorepo.
The biggest problem we currently have, that triggers all the discussions, is that we try to be aligned with one of the best practices, which is "build artifact once - deploy it to any env at any time". If the source code is changed the process is quite simple and straight forward. Build new artifact, deploy to stage then deploy to prod. But the problem emerges when helm chart was changed but the source code not; the previous version of artifact must be deployed. The best solution would be the cheapest solution and the running-fast one. We can not agree on how to get that "previous version". Very much possible that if we delegate deployment task to GitOps tool it would solve the problem, but as I mentioned above we can not do such separation right now. Another option is to use some kind of external storage or DB from which we can retrieve the artifact version, but external storage leads to additional expenses and possibly complected custom solution. Or we can try to mimic GitOps tool behavior by committing version directly to helm chart values file of Chart.yaml or template - doesn't matter.
1. Have you had similar problem?
2. How did you solve it?
3. What in your opinion we do wrong and how would you tackle the problem?
4. Is there any solution which is bitbucket native?
