Linux Container failing after Network Drive SMB unmount

First I have to ensure a few configurations are set in Unraid

• Unraid has a static address (I already implemented), so the IP doesn't change
• Stop disks from spinning down on the disks backing the share
• Disable SMB Multichannel
• Ensure the Proxmox user is configured correctly and has read/write access

Next I have to store the SMB share credentials on node 1.

I just have to enter the shell and input

cat > /root/.smbcred <<'EOF'
username=proxmox-node1
password=**
domain=WORKGROUP
EOF
chmod 600 /root/.smbcred

Creating Directory

First, I have to create the share directory
mkdir -p /mnt/unraid-public-storeage

Hard Mounting

We add the lines above into /etc/fstab

mkdir -p /mnt/unraid-public-storage

cat >> /etc/fstab <<'EOF'

# Unraid public storage
//192.168.1.240/PublicStorage /mnt/unraid-public-storage cifs credentials=/root/.smbcred,vers=3.1.1,hard,echo_interval=10,noserverino,uid=100000,gid=100000,file_mode=0770,dir_mode=0770,_netdev,nofail,noauto,x-systemd.automount,x-systemd.mount-timeout=30 0 0
EOF

##Then to confirm it worked
tail -5 /etc/fstab
systemctl daemon-reload
systemctl restart remote-fs.target
ls /mnt/unraid-public-storage

What this does is it creates a hard mount and probes the connection every 10 seconds, gives it the right privileges, and waits for the network to remount, allowing that task to continue. I also added nofail so this wouldn't stop Proxmox from booting.

Then we can run the command below to mount our new hard-mounted, fail-proof share to our media Docker container

pct stop 101
pct set 101 -mp0 /mnt/unraid-public-storage/Media,mp=/mnt/unraid-media
pct start 101
pct exec 101 -- ls /mnt/unraid-media

Since the folder share is not mounted to the CT, we have to pass it through Docker so our media share can read and modify the share

It would still be the same process as [[mounting-network-shares-on-my-home-lab-nfs-vs-smb#mounting-to-docker-container | as this]

All I have to do is modify my stack in Portainer and rebuild it

Post-reboot, ensure that the service can see the folder

All we're doing is rebooting our media container and checking whether we can still see the contents of the share directory

From the Node Shell

pct stop 101
pct start 101

On reboot from the CT shell

We're going to try to just cd to the share location

Success

This is a realistic scenario that might occur when I'm running maintenance on Node 2

To complete this test, I'm going to use the script my LLM made to test the mount connection every ten seconds on port 445 (TCP)

while true; do timeout 3 bash -c ">/dev/tcp/192.168.1.240/445" 2>/dev/null && P=up || P=DOWN; findmnt -n /mnt/unraid-media >/dev/null 2>&1 && M=mounted || M=MISSING; timeout 5 stat -f /mnt/unraid-media >/dev/null 2>&1; rc=$?; case $rc in 0) I=ok;; 124) I=BLOCKED;; *) I=ERR$rc;; esac; echo "$(date +%T) 445=$P mount=$M io=$I"; sleep 10; done

Testing

First, I'm going to run the script and monitor it

This output indicates the share is mounted and connected. Now I will stop my array and see the results

Now we're getting an I/O error, as expected. The real test is to restart the array and ensure the mount works as normal and that the services still see the share.

Even after restarting the mount, our script logged the disconnect and the reconnect.

Now all I have to do is ensure my services are still connected, and this test is complete.

Success

This time around, everything remounted automatically and my services picked up where they left off.

This test is mandatory because, in the last test, even though the array was off, the Node could still send a rejection when the CT sent a port 445 TCP request. In a scenario where the share is unexpectedly powered down, such as node-2-overheating-issue, I want to ensure on reboot the CT automatically remounts as if nothing ever happened. This was the issue this incident fix never resolved.

I'm going to run the same script from the last test, reboot my node, and test the results.

We got our typical error due to the share not being available; it still shows as mounted, which means our hard-mount implementation is working as intended. Now we just wait for Node 2 to finish rebooting.

It takes a few minutes for my array to fully start, but

Successful