On truenas-core 12-u6 the zfs-section is not properly working. I do get some basic metrics (see below), but they do not make it to the influxdb (zfs_pool and zfs_dataset did not even got created for this host, so this metric has no entries), and are therefore not useable within grafana. other hosts with zfs-pools work fine, but these are debian hosts (deb9/10 with proxmox 6/7). The relevant config-snippets are identical accross systems, so I assume something in the output breaks some filtering logic.
I'd assume, that this might come from an awkward named pool or dataset. If I would use silly names with special characters, i would deal with is, but I assume the truenas-default pools might not be correctly pased, because on a second, similarily configured truenas-instance with only one other pool named backup, the behaviour is exactly the same: no dataset or pool-metrics, so this would very likely affect everyone else also.
Relevant telegraf.conf:
[[inputs.zfs]]
poolMetrics = true
datasetMetrics = true
System info:
Telegraf-Version (packaged binary from github, sha256: d87514f06cd2efa6990383906d3b4664cb17222fde6755442b653a13a42c201a )
Telegraf 1.20.2 (git: HEAD f721f53d)
Kernel: FreeBSD 12.2-RELEASE-p10 b26f74b5984(HEAD) TRUENAS amd64
zfs:
zfs-2.0.5-1
zfs-kmod-v2021071201-zfs_f7ba541d64cbc60b21507bd7781331bea1abb12e
Expected behavior:
zfs-plugin collecting dataset and pool metrics
Actual behavior:
only very limited basic-metrics
Additional info:
I attached a sample output of the questionable machine
./telegraf --config telegraf.conf --input-filter zfs --test
telegraf-zfs.txt
ultimately i don't know, if this is an influx(1) or telegraf-issue, but as i said, the same setup works find on linux.
On truenas-core 12-u6 the zfs-section is not properly working. I do get some basic metrics (see below), but they do not make it to the influxdb (
zfs_poolandzfs_datasetdid not even got created for this host, so this metric has no entries), and are therefore not useable within grafana. other hosts with zfs-pools work fine, but these are debian hosts (deb9/10 with proxmox 6/7). The relevant config-snippets are identical accross systems, so I assume something in the output breaks some filtering logic.I'd assume, that this might come from an awkward named pool or dataset. If I would use silly names with special characters, i would deal with is, but I assume the truenas-default pools might not be correctly pased, because on a second, similarily configured truenas-instance with only one other pool named
backup, the behaviour is exactly the same: no dataset or pool-metrics, so this would very likely affect everyone else also.Relevant telegraf.conf:
System info:
Expected behavior:
zfs-plugin collecting dataset and pool metrics
Actual behavior:
only very limited basic-metrics
Additional info:
I attached a sample output of the questionable machine
./telegraf --config telegraf.conf --input-filter zfs --testtelegraf-zfs.txt
ultimately i don't know, if this is an influx(1) or telegraf-issue, but as i said, the same setup works find on linux.