@mhamzahkhan I have looked I to it, but my bigger concern is that there is no way back >.> means once migrated I'm required to run K8s or at least can't re-unify the data endpoint such as artifacts and alike.
@sheogorath Oh you mean because the data is stored in S3 (or some other similar storage provider)?
I'm actually already storing artifacts, and everything that I can in S3 ("uploads", artifacts, registry, and LFS).
The bit I am confused about is how to handle the actual repository data.
@sheogorath From my understanding the repo data can't be stored in S3, but rather the recommended way is to create a Gitaly Cluster with some form of storage (probably local storage), and Gitaly replicates the repos across all the gitaly nodes.
I'm a bit stuck with that because my K8S cluster only has NFS as a PV provider, and NFS isn't recommended.
@mhamzahkhan yes, repodata is storend on a PVC. But I also wouldn't say it has to be super super fast, so NFS might be fast enough. Maybe try it without performing a full migration but just setting up a test instance?
@mhamzahkhan well, integrity concerns are valid in general, because NFS does local caching and data considered written to disk might not have made it to the NFS server yet. But honestly speaking, as long as your host machine isn't crashing, that should be fine.
Latency can be considered a problem for something like a database or elasticsearch. There is not much you can do about that. For the plain git storage however right now, I see no problem with NFS.
This is my personal microblog. It's filled with my fun, joy and silliness.