ZFS DDT metrics

[jpds]

The exporter should grab the ZFS DDT metrics from: /proc/spl/kstat/zfs/$poolname/ddt_stats_*

[discordianfish]

Please provide a bit more details What are these? Why do you need them etc. Just so we have a bit more context to decide if we want them without having to do that research ourselves. Thanks!

[jpds]

It's the stats for the ZFS deduplication table. Now that we have https://klarasystems.com/articles/introducing-openzfs-fast-dedup/ we have a filesystem that can efficiently dedup the same data across filesystem blocks. Very useful for things like Docker or virtual machines hosts.

[discordianfish]

Cool ok, I see this being useful. As usual, the parsing needs to go into prometheus/procfs and then can be used here.

[michael-duren]

@jpds is this being worked on?

[michael-duren]

I'd like to take a stab at it, if no one else is working on it.

[natneam]

Hi @michael-duren, are you still working on this? If not, I'd like to pick it up. I already have a working implementation and test fixtures captured from a real OpenZFS 2.3 system.

Per the earlier comment, my plan is to:

1. Add a parser for /proc/spl/kstat/zfs/<pool>/ddt_stats_<checksum> to prometheus/procfs.
2. Once that's merged, use it in the zfs collector here to expose node_zfs_zpool_ddt_* metrics with zpool and checksum labels.

If you've already started, happy to step back or collaborate. Otherwise I'll open the procfs PR in the next few days.