Closed kylehuntsman closed 2 years ago
I believe @masih you were saying there may be some machine reservation stuff as well here because the worker nodes currently don't have this much memory.
- This should move to the dev overlay as a patch.
@masih Why did moving to a larger instance type not require this change?
Because that suggestion is about keeping resources out of the base kustomization layer; it's a best practice, cause we don't know who might deploy autoretrieve and with what config. The resources in the base layer, then, might be an over or under utilisation :)
Separation of manifests is totally unrelated to instance types.
Autoretrieve is still running out of memory. Bumping the memory request to 32GiB, the size of a sector. The idea being a CID being requested might be the entire sector, so hopefully this is a high enough limit that it doesn't fall over as often.