There's a race condition. It doesn't recover after additional puppet-runs as far as I know. Just requires a manual reboot of the instance.
Happens to about 3 out of 10 instances.
Description
Description
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | • Bstorm | T126083 overhaul labstore setup [tracking] | |||
| Resolved | None | T91822 New instances don't always get their /data and /home project mounted from NFS |
Event Timeline
Comment Actions
The logic there now is showmounts waits for the NFS export to be present on initial puppet run if an instance is meant to have NFS mounts. This waits for the nfs-exports task to run on labstore1001 to fulfill async to this initial VM up operation. That server has historically been way overloaded, but with some of our recent changes we are surviving in lieu of more hardware and I am unable to reproduce this atm. I am resolving until we see it again.