Post #1598036
2026-04-23 21:42 UTC
@mmeier@social.mei-home.net @louis@social.louis-vallat.dev there are many ways to somewhat easily provision a RWX these days (nfs subdir prof, longhorn, nfs ganesha, etc.).
A nodeSelector, as previously mentioned, might also do the job.
FWIW , IMO running a cluster without any RWX option is somewhat shortsighted.
Replies (1)
-
@mmeier@social.mei-home.net 2026-04-23 22:05
@pat_dev@social.linux.pizza My issue isn't so much that I don't have RWX volumes. I do, based on CephFS. The issue here is two-fold, and has to do with OCI image builds. CephFS is slower than using an RWO RBD as the basis for the workflow volumes. Plus, OverlayFS does not work on CephFS, so I'd have to use the slower VFS driver. Both would result in far longer image builds. And Woodpecker doesn't allow configuring the StorageClass to be used per workflow, only for the entire instance. @louis@social.louis-vallat.dev