- Shipped
- November 4, 2026 at 4:57 PM UTC
- Author
- Kamo
- Commit
- ccfdec0
The worldchunks USB drive dropped off k3m1 on 2026-10-30 and has not come back, so nfs-server stopped. k1m1 kept its soft mount on the assumption the server returns quickly; for five days every statfs on it blocked ~42s. node-exporter's filesystem collector then blew Prometheus's 10s scrape timeout on every other scrape, leaving k1m1 node metrics and alerts blind. Now the client lazily unmounts when :2049 does not answer, and the existing branch mounts it again once the server is back. Nothing on k1m1 reads it.