Free space mismatch

[lazaroblanc]

The zfsprom_free metric is showing over 200 G free, while zfs list shows 67 G free.
Any idea how this happens?

[matusnovak]

Hi.

This is interesting. I can not reproduce it on my pool.

The calculation is implemented as:

# "stats" are from the libzfs library
metric_free = stats.size - stats.allocated

But I believe this is a wrong calculation.

Could you share the output of your zpool list command of pool1? And do you have any snapshots?

Also, FYI, I would recommend using zpool_influxdb instead. It has been integrated into the ZFS apt package. You can read InfluxDB formatted metrics (InfluxDB Telegraf can read them and convert to Prometheus formatted metrics).

[lazaroblanc]

Ah that's interesting. Here it shows 204 G "FREE" but before with zfs list it shows 67 G "AVAIL"

AVAIL would be the correct metric since that's actually usable space. FREE shows the entirety of free space including reserved "slop space"

[lazaroblanc]

Also, FYI, I would recommend using zpool_influxdb instead. It has been integrated into the ZFS apt package. You can read InfluxDB formatted metrics (InfluxDB Telegraf can read them and convert to Prometheus formatted metrics).

Thanks for the recommendation but I liked this script because of the simplicity of the setup. If I migrate I might just do it with a bash oneliner or script and the node_exporter text collecter since that's already running anyway :)

[matusnovak]

I believe I can add a metric for AVAIL.

However, due to truenas/py-libzfs#249 I am no longer able to build a new Docker image.

This issue is blocked until then :(