Summary

Fixes: #21518

Problem

When a ZFS pool runs out of disk space, the diskspace collector triggers separate alerts for every dataset in the pool. For pools with hundreds of datasets, this creates massive alert/notification floods.

Root Cause

The diskspace collector treats each ZFS dataset as a separate mountpoint and calls stat() on every one. When the parent pool fills up, all child datasets are affected simultaneously, triggering alerts for each one.

Solution

Exclude ZFS datasets from diskspace monitoring, keeping only ZFS pool-level mountpoints. This is done by checking the mount source field in /proc/self/mountinfo:

• ZFS pools have no slash in the source name (e.g., tank)
• ZFS datasets contain slashes (e.g., tank/install, tank/u/user1)

Test Plan

Additional Information

Summary by cubic

Prevents alert floods in diskspace monitoring by excluding ZFS datasets and monitoring only ZFS pools.

• Bug Fixes
  • Skip ZFS datasets detected via mount_source containing a slash in /proc/self/mountinfo (e.g., tank/install).
  • Continue monitoring pool-level mountpoints without a slash (e.g., tank) so alerts fire once per pool.

^{Written for commit 0c0ff65. Summary will update on new commits.}