ReplicateRepository always fails after 10 minutes
We are seeing this in production now, which may be explain the ever-growing Praefect replication queue for praefect-03
: https://log.gprd.gitlab.net/goto/202f8c50-7d77-11ec-a649-b7cbb8e4f62e
SSHUploadPack
has a hard-coded 10-minute limit here: https://gitlab.com/gitlab-org/gitaly/blob/400c7225dd2a0dd17e907ce5af25e7ade5ed81fb/internal/gitaly/service/ssh/server.go#L15.
Maybe we want to consider raising this limit? Or raise it if the client if it's from a ReplicateRepository
?
I suspect it's taking this long in order to synchronize the millions of keep-around refs: https://gitlab.com/gitlab-org/gitaly/-/issues/2436#note_820297158
Edited by Stan Hu