Skip to main content
RunBook Academy

← All break/fix scenarios in Proxmox VE

intermediateNetworking~15 min

NFS share becomes read-only during heavy writes

Reported symptoms

  • VMs on NFS-backed storage show I/O errors during writes
  • NFS client reports "Stale file handle" or "Input/output error"
  • NFS server exports show the client as "read-only"
  • Backup or migration to NFS fails

Evidence

  • · mount | grep nfs (look for "ro,")
  • · cat /var/lib/nfs/rmtab on the NFS server (or equivalent)
  • · dmesg on the NFS client shows "nfs: server not responding"
  • · cat /proc/mounts shows the NFS mount as ro
Diagnosis and resolutionclick to reveal

Root cause

The NFS server or client has marked the mount as read-only. Common causes: - NFS server was reconfigured with `ro` option - Kernel detected I/O issues and remounted read-only to protect data - NFS server exceeded max client connections and rejected new ops - Firewall blocking the mount-up protocol

Remediation

1. Check the NFS server exports: `cat /etc/exports` on the NFS server `exportfs -v` (verify what's actually exported) 2. Check the mount on the client: `mount | grep <mount-point>` `cat /proc/mounts | grep <mount-point>` 3. If read-only, try remounting: `umount <mount-point>` (if possible without disrupting VMs) `mount -o rw,vers=4.1 <nfs-server>:/path <mount-point>` 4. If the client cannot unmount (VMs are running), force remount: `mount -o remount,rw <mount-point>` 5. If the server marked the client read-only due to too many operations, lower NFS tuneables: On the server, increase: `/proc/fs/nfsd/max_write_size` `/proc/sys/net/core/wmem_max` 6. Check server logs: `/var/log/syslog` (look for nfsd errors)

Verification

- `mount | grep nfs` shows "rw," - VMs on the NFS share respond normally - Backup and migration succeed - No new I/O errors in dmesg

Prevention

- Use NFS v4.1 or later (better semantics) - Configure adequate server buffer sizes - Monitor NFS mount state with a check - Document the NFS server tuning parameters