Defining the mountpoints per cluster for the filesystem metrics

[dex4er]

Fixes #145

[dex4er]

I'm not sure if we should leave / as default or set it to .*

@robertobandini Do you have any AKS for verification maybe? I tested it on EKS with Ubuntu and Bottlerocket OS.

Before:

Ubuntu

Bottlerocket

After:

Ubuntu

Bottlerocket

[dex4er]

Ok, so the finding is that / is too less and .* is too much. node-exporter duplicates some filesystems if they are mounted many times then numbers are summed incorrectly. In my case it should be ~1TB but the same filesystem is bind-mounted 4 times.

The safe default would be /|/local|/mnt|/mnt/.*|/var which should cover the most important systems (Bottlerocket, COS, Talos. Ubuntu)

Edit:

Uh, /var and /mnt are doubled on Bottlerocket too then we won't have universal regexp for every OS

[robertobandini]

Uh, /var and /mnt are doubled on Bottlerocket too then we won't have universal regexp for every OS

Ok, so If I understood, we need a different value depending on the OS, isn't it?
Maybe we can autodetect it?
At first glance in summary on AKS the old default value works on Linux nodes, but if needed I can do an in-depth test.

Note: to see the disk metrics I had to put Helm instead of Auto detect; it seems to me that something needs to be revised regarding auto detect, what do you think?

Anyway: excellent and fair proposal and excellent work!

[dex4er]

Maybe it should be heuristic for the default value per different OS like ie. for node shell. Something is better than nothing then I left just the single value per cluster to set.