Have nfs share setup to automount in a laptop client. The server occasionally goes offline, and when it does, the laptop cannot reboot or shutdown (it seems to get stuck looking for the share). Are there mount options that tell the client to ignore missing share and proceed to reboot or shutdown cleanly, without having to manually umount the share.
The automount is performed by systemd mount and automount unit files. these are the relevant mount options currently used:
can you supply the systemd unit mount file for the volume ?
is the mount in /etc/fstab ? if so, can you copy/paste the entry.
are there any (systemd) dependencies on the volume being mounted before continuing.
If you were asked the question - what details would you expect to be given ?
the more details you give the less questions need asking and better the chance of .. a solution / questions that need asking
sure. they are automount and mount units not fstab as stated in opening post. for
simplicity sake, the location of the export directory on the server is the same as the mount point of the client.
TimeoutIdleSec=
Configures an idle timeout. Once the mount has been idle for the specified time, systemd will attempt to unmount. Takes a unit-less value in seconds, or a time span value such as "5min 20s". Pass 0 to disable the timeout logic. The timeout is disabled by default.
perhaps removing that may have some influence on behaviour.
This might work, depending on the kernel version and depending on the state of a current access. Some fuser implementations do a stat() and/or statfs() that gets hung itself...
Linux has got the "lazy" umount umount -l /mnt/Data
that succeeds in nearly all access states, but leaves some dead resources in RAM. But it's perfect if run before a reboot/shutdown!
Last but not least, the mount options.
There is the classic soft mount option that after a time-out sets all accessing file handles to "I/O error". Linux has got some alternate "soft" options that you can try. See man nfs
or https://man7.org/linux/man-pages/man5/nfs.5.html
in a script that another systemd service unit can run prior to reboot/shutdown? i had been considering that earlier. was hoping that some mount option that i had overlooked could do the job without complicating the matter.
i will try out soft option, all my files are relatively small, they should not get lost in transit.
update: the soft option does not do it. next i will look at alternate soft options.
Put it in a script, followed by the umount -l /mnt/Data
and run the script on the command line when the nfs server is not reachable.
In case the fuser are hung for a long time, disable them, and rerun the script with the remaining umount commands.
If even the umount -f is hung for a long time then disable that, too.
in the time i took to make the suggested script, i realized that using systemd automount/mount + a script to mimic what autofs does in other distros (broken/unmaintained in my laptop clearlinux) is sorely inadequate.
i asked bing chat ai to write a script that
mounts the share when i start laptop;
monitors my server;
mounts the share if server is online;
clean up the stale mount if server is offline (so nothing hangs when i shutdown or reboot laptop);
remount the share if server comes online again;
unmounts share after amount of time of no activity;
unmounts share prior to reboot/shutdown
i have systemd service to start script.
on the surface, the script seems to be working, but i have not test all scenerios.
one problem i noticed so far is when i have the share open in nautilus file manager, it goes offline after a certain time - due to inactivity? i want it to unmount only if i am not accessing the share ie: i do not have the share folder open.
i am not very computer literate, hoping extra sets of eyes can help:
#!/bin/bash
# Define the mount point and the remote server details
MOUNT_POINT="/mnt/data"
REMOTE_SERVER="server"
REMOTE_PATH="/data"
INACTIVE_TIME=300
# Function to check if the remote server is online
function is_server_online {
ping -c 1 $REMOTE_SERVER > /dev/null 2>&1
return $?
}
# Function to clean up stale mount point
function cleanup_stale_mount {
# Check if the mount point is already mounted
if mountpoint -q $MOUNT_POINT; then
# Unmount the stale mount point
sudo umount -f -l $MOUNT_POINT
# Check if the unmount was successful
if [ $? -eq 0 ]; then
echo "Successfully unmounted stale mount point $MOUNT_POINT"
else
echo "Failed to unmount stale mount point $MOUNT_POINT"
fi
fi
}
# Function to unmount the share when not in use
function unmount_share {
# Check if the mount point is already mounted
if mountpoint -q $MOUNT_POINT; then
# Check if any processes are accessing the mount point
if fuser -m $MOUNT_POINT > /dev/null 2>&1; then
echo "$MOUNT_POINT is in use, not unmounting."
sleep $INACTIVE_TIME && unmount_share &
else
# Unmount the share
sudo umount $MOUNT_POINT
# Check if the unmount was successful
if [ $? -eq 0 ]; then
echo "Successfully unmounted $MOUNT_POINT"
else
echo "Failed to unmount $MOUNT_POINT"
fi
fi
fi
}
# Function to mount the share when server is online and share is not mounted.
function mount_share {
# Check if the remote server is online and mount point is not already mounted.
if is_server_online && ! mountpoint -q $MOUNT_POINT; then
# Create the mount point if it doesn't exist.
if [ ! -d $MOUNT_POINT ]; then
mkdir -p $MOUNT_POINT
fi
# Mount the remote path to the mount point.
sudo mount -t nfs $REMOTE_SERVER:$REMOTE_PATH $MOUNT_POINT
# Check if the mount was successful.
if [ $? -eq 0 ]; then
echo "Successfully mounted $REMOTE_SERVER:$REMOTE_PATH to $MOUNT_POINT"
sleep $INACTIVE_TIME && unmount_share &
else
echo "Failed to mount $REMOTE_SERVER:$REMOTE_PATH to $MOUNT_POINT"
fi
elif ! is_server_online; then
cleanup_stale_mount
fi
sleep 10 && mount_share &
}
# Start monitoring for mounting and unmounting.
mount_share &
# Function to handle shutdown and reboot signals.
function shutdown {
cleanup_stale_mount
exit 0
}
trap shutdown SIGINT SIGTERM SIGKILL
# Keep script running until shutdown or reboot signal is received.
while true; do sleep 1; done;
Hmm, it is all plausible.
I suspect the background things, because you never know when they are finished.
One is a umount attempt after 300 seconds, but what is this: sleep 10 && mount_share &
This is within the mount_share function - does it call itself after 10 seconds??
Regarding the combined umount -f -l I have no experience. I have always used umount then umount -f (can leave resources behind on the NFS server) then umount -l (can leave local resources behind).
this is how ai explains the issue with inactivity even though the share is mounted to nautilus:
If you have a Nautilus window open on the mount point directory but are not actively working with any of the files, the fuser command may or may not detect this as activity, depending on how Nautilus interacts with the file system. If Nautilus keeps an open file handle on the directory or any of its files, the fuser command will detect this as activity and the script will not unmount the share. If Nautilus does not keep any open file handles, the fuser command will not detect any activity and the script will unmount the share after the specified period of inactivity.
regarding sleep 10, it is suppose to ping the server every 10sec.
i am not quite there by a long shot, there is alot to mull over. also, i cannot edit the title anymore to better reflect what i need, which is basically autofs mimicry.
not sure exactly how to monitor shell processes, but if this helps, htop looks normal even after a few hours of clicking in and out of the share.
tried the soft option previously with the mount unit file, the system (laptop) was still hanging at reboot and shutdown when the server had gone offline.
and better the chance of .. a solution / questions that need asking