vpenso/prometheus-slurm-exporter

no metrics

Closed this issue · 3 comments

hoba87 commented

I've followed the DEVELOPMENT.md for installation, but when trying to use the command curl http://localhost:9103/metrics after ./bin/prometheus-slurm-exporter --listen-address="0.0.0.0:9103", there is no output and the curl command waits until I kill the exporter.
The output from the exporter is only:
INFO[0000] Starting Server: 0.0.0.0:9103 source="main.go:48"

Any idea why? I use CentOS 8.2 with Prometheus 2.22.0 and Slurm 20.02.5.

mtds commented

I need a little bit more information:

  1. Are you launching the curl command on the same host or VM where the exporter is running?
  2. Is the exporter running where sinfo, squeue etc. are properly installed and able to query your Slurm instance?
  3. Do you have firewall policies (e.g. implemented via firewalld ) which limit/restrict network connections?
hoba87 commented

Are you launching the curl command on the same host or VM where the exporter is running?

yes, on the same host

Is the exporter running where sinfo, squeue etc. are properly installed and able to query your Slurm instance?

yes, the path of slurms bin directory is added to the PATH variable

Do you have firewall policies (e.g. implemented via firewalld ) which limit/restrict network connections?

firewalld is installed but only used for external network, however I have tried it with temporarily disabling the firewall, with the same result.

hoba87 commented

I think the problem was using Slurm with Accouting to text file, because an sacct run takes very long. I have switched to slurmdbd now, with no problem anymore.

  • 👍1