Proxmox reporting empty disk reads/writes for LXC containers

There seem to be mixed issues in this thread
Click to expand...

I agree it's maybe a bit harder to follow. I would suggest trying to replicate these variants:
- docker in LXC making writes (I think you already confirmed that)
- LXC with no docker in, rootfs on LVM (or whatever) and data disk on ZFS
 
I have 4x LXC containers that are not reporting DiskIO stats, even when using "pvesh get /nodes/localhost/lxc/#####/rrddata --timeframe hour" it reports no read/write IO.

There is a pattern though. This started when the Proxmox VE node abruptly was shut down (for reasons I am not getting into) and then turned on later. Also all 4x of my LXC contianers _only_ run on that one node, not any other node. And the PVE v7.x->8.x upgrade was performed multiple months before-hand and the graphs were working for these containers after that upgrade for multiple months.

So in my case I'm suspecting my issue could be "fixed" when we update/reboot this PVE Node next (whenever that is). And this might be the case for some other people in this thread. Considering how LXC hooks into the Linux Kernel there could be some "busted" hooks for these 4x containers that have stayed "bad" this whole time.

Can't find much evidence otherwise. Restarted pvestatd and lots of other recommendations in this thread and others, no change.

For your consideration.
 
Okay, so.

I looked into apparently everything that was written here, and nothing seems to change the situation, as in, there are still no disk I/O metrics.

Rebooting also doesn't help, as I restarted all hosts recently due to 8.2 upgrade, and this also didn't fix the issue.

What else can I do to debug it?
 
  • Like
Reactions: BloodyIron
having the same issue here, multiple pve hosts. (v8.4.1 and v8.3.3).

some LXC's do report stats, some LXC's dont.

ran a test, created a fresh ct; ubuntu 24.02 everything default.
downloaded a 500MB file.
after 24h, no writes reported. Not on the dashboard, not on the metric server output (grafana).

After days of comparing and trying settings in the conf i gave up.

did some searching and made a workaround;
Code: 
cat /proc/$PID/io
does report the writes.

after some copy pasting in bash i now have (cumulative totals (raw values get written to a file ))
1749731464395.png
next is to get this into the metrics server.
 
Last edited:
You must log in or register to reply here.
Top Bottom