System integration does not report filesystem metrics for network-mounted filesystem

Elastic Stack Elastic Agent
1

Hi,

I’m collecting host metrics from my Elasticsearch nodes using the System integration.

Filesystem metrics are collected correctly for all of my local mount points under /mnt/*, for example /mnt/es_data, but I’m not getting any metrics at all for /mnt/other_data.

The only difference I can see is that /mnt/other_data is an nfs4 mount whereas the others are local filesystems.

I’ve already verified that the mount is present and accessible (findmnt, df -T, /proc/mounts and /proc/self/mountinfo all show it correctly), but in Kibana, i still do not find anything in system.filesystem.mount_point for /mnt/other_data data stream directly.

I also checked the System integration configuration in Fleet but couldn’t find anything related to excluding NFS or network filesystems.

Is this expected behaviour for the System integration, or should nfs4 filesystems be collected as well?

If this is expected, what is the recommended way to collect filesystem metrics for NFS-mounted filesystems? If not, is there something else I should be checking?

Environment:

  • Elastic Stack: 9.3.3
  • System integration: 2.16.1
  • Elastic Agent: 9.3.3 and 8.19.9
  • RHEL 9.8

Thanks

2

Hi Barbarossa. This could be Metricbeat’s default filtering. When filesystem.ignore_types is omitted on Linux, Metricbeat ignores filesystem types marked nodev in /proc/filesystems, which can include nfs4.

In Fleet, can you try to edit the System integration and configure List of filesystem types to ignore under both System filesystem metrics and System fsstat metrics so the rendered Agent policy contains filesystem.ignore_types: []?

In other words, are both of these empty?

Screenshot 2026-08-19 at 3.22.39 PM
Screenshot 2026-08-19 at 3.22.39 PM1486×1290 82.6 KB

3

Hi,

Thanks for your answer.

I checked this and confirmed that nfs4 is not configured to be ignored under either the filesystem or fsstat metrics. I also checked the rendered/internal Elastic Agent configuration and couldn’t find anything indicating that nfs4 is being excluded there.

What makes this more interesting is that the same Elastic Agent, using the same System integration, is successfully collecting filesystem metrics for another nfs4 filesystem. The issue only occurs with this particular network-mounted filesystem.

So my current suspicion is that the problem is related to the way this specific network filesystem is mounted or handled, rather than nfs4 being excluded in general.

If you have the possibility to reproduce this with a network-mounted NFS filesystem on your side, it would be interesting to know whether you see the same behavior.

In the meantime, I’ve found a workaround for monitoring this filesystem, so from my side this topic can be considered closed for now.

Thanks again for your help!