Proxmox VE Memory Usage CRIT but Memory OK

Hello everybody,

we are running a proxmox cluster and we installed a Checkmk Enterprise Edition 2.1.0p23. We follow the instructions of Proxmox Monitoring: How to Do it Efficiently with Checkmk | Checkmk and everything works as designed. But my problem is that the service Proxmox VE Memory Usage on the VMs (Ubuntu OS) show CRITs and WARNs but the memory service is OK because the OS use RAM for caching and buffering. So whats the best practice for those services Proxmox VE Memory Usage? Stop notifications, change the levels?

Thx in advance and kind regards,
Frank

1 Like

If the check’s default parameters do not fit your situation, you can change them.

1 Like

I, too, am puzzled by the CRIT (730%) memory. It is only a single VM running HAPROXY on Debian Buster. It idles and uses virtually no resources at all and no matter how much RAM/SWAP I throw on it, the error never goes away.

I have dozens of VMs in the cluster and only one is throwing this puzzling CRIT.

Perhaps some of us are misreading the numbers below, but there is just no way I can make them add up to 730% committed.

If you have a suggestion of what I might adjust in the parameters, I’d love some input.

Total virtual memory: 2.92% - 119.43 MB of 4.00 GB, Committed: 730% - 29.23 GB of 4.00 GB virtual memory (warn/crit at 100.00%/150.00% used)**CRIT**, 8 additional details available
Details Total virtual memory: 2.92% - 119.43 MB of 4.00 GB
Committed: 730% - 29.23 GB of 4.00 GB virtual memory (warn/crit at 100.00%/150.00% used)**CRIT**
RAM: 2.69% - 55.13 MB of 2.00 GB
Swap: 3.14% - 64.30 MB of 2.00 GB
Commit Limit: -391.0% - -15.67 GB of 4.00 GB virtual memory
Shared memory: 3.43% - 70.27 MB of 2.00 GB RAM
Page tables: 3.96% - 81.07 MB of 2.00 GB RAM
Disk Writeback: 0% - 0.00 B of 2.00 GB RAM
RAM available: 97.31% free - 1.95 GB of 2.00 GB
Hardware Corrupted: 0% - 0.00 B of 2.00 GB RAM

This may not be OP’s exact situation, but it may be related.

It’s related as it is another memory check, but from inside the VM.

What you see is a high amount of committed memory. This is memory that is allocated by processes but not used yet, i.e. not written to. Some applications just malloc() large amounts of memory without using it. But if they would write to it, RAM and swap could be overused.

You can set thresholds for committed memory in this service check or even just ignore it if your application runs fine.

1 Like

thanks @r.sander, I think you may be right. There may be a HAPROXY setting that would fix this as well.
Seems wierd, but the server RAM idles at 2% usage while throwing this 700 - 900% CRIT on committed memory. Since it’s in a kvm, RAM will expand as required anyhow so I just wrote a rule to Warn:900%:Crit:1000%.
Strangest rule I’ve ever written though.