zlacelle / nagios_check_zfs_linux Goto Github PK

Looks good so far.

Two minor issues:

1. error message if sudo is missing. If sudo is not installed the script fails with "invalid pool"
2. better locale support. The script fails if run with LANG=de_DE (e.g. comma as decimal delimiter) the script fails to parse some numbers:

Traceback (most recent call last):
  File "./check_zfs.py", line 260, in <module>
    sizeGB = ConvertToGB(size)
  File "./check_zfs.py", line 74, in ConvertToGB
    gigs=float(value)*1024
ValueError: invalid literal for float(): 8,12

Without further knowledge of the intended scope of this script, here are some ideas for additional checks regarding the health of the zpools:

• scrub status (there exists a project that does that already as shell script: https://github.com/Josef-Friedrich/check_zpool_scrub)
• disk status
• disk error counts (READ/WRITE/CKSUM)
• data errors

So basically it would be nice to get a warning if zpool status shows anything unusual. (I admit that might be hard to generalize, I think I try a simple regex solution first.)

The code checks for Python 3.5, but that version actually does not work, as the f-strings are used, which require Python 3.6. This effectively drops support for Ubuntu 16.04 (which has reached "end of standard support").

There are only 3 places where f-strings are used, which could be replaced, otherwise the version check should probably test for Python 3.6.

Hi,
I noticed that on Proxmox all users can run zfs commands. Is that also the case for other systems? Are there any reasons to avoid running the commands as an unprivileged user?

If sudo is required, I would also suggest being more restrictive in the suggestions you include in README. Something like this seems to be enough and doesn't grant access to any dangerous zfs commands:

<user> ALL=(ALL) NOPASSWD: /sbin/zfs list
<user> ALL=(ALL) NOPASSWD: /sbin/zpool list <pool>
<user> ALL=(ALL) NOPASSWD: /sbin/zfs get compression <pool>
<user> ALL=(ALL) NOPASSWD: /sbin/zfs get compressratio <pool>

are there any plans to support an argument to check any zpools present? the use case i'm concerned about handling is adding a pool to a system, monitoring not being updated immediately, and losing a disk. being able to run check_zfs --allpools means any zpools present will report degradation, even if the pools are brand new, inherited during failover, etc.

thanks for the plugin!

Running the script with python3 yields errors, probably due to minor things like print statements.
Since python2 is not installed by default anymore on modern distros, and will soon be EOL, this would benefit from being ported to python 3.