modules/nixos: add cgroups and exporter #1466
Conversation
flake.nix
Outdated
| @@ -16,6 +16,8 @@ | |||
| buildbot-nix.inputs.nixpkgs.follows = "nixpkgs"; | |||
| buildbot-nix.inputs.treefmt-nix.follows = "treefmt-nix"; | |||
| buildbot-nix.url = "github:qowoz/buildbot-nix/more-skipped-build-workers"; | |||
| cgroup-exporter.inputs.nixpkgs.follows = "nixpkgs"; | |||
| cgroup-exporter.url = "github:arianvp/cgroup-exporter"; | |||
There was a problem hiding this comment.
Still need to send a PR upstream for aarch64-linux support.
flake.nix
Outdated
| @@ -16,6 +16,8 @@ | |||
| buildbot-nix.inputs.nixpkgs.follows = "nixpkgs"; | |||
| buildbot-nix.inputs.treefmt-nix.follows = "treefmt-nix"; | |||
| buildbot-nix.url = "github:qowoz/buildbot-nix/more-skipped-build-workers"; | |||
| cgroup-exporter.inputs.nixpkgs.follows = "nixpkgs"; | |||
| cgroup-exporter.url = "github:arianvp/cgroup-exporter"; | |||
There was a problem hiding this comment.
What is the difference to this plugin? https://github.com/influxdata/telegraf/tree/master/plugins/inputs/cgroup
There was a problem hiding this comment.
Oops, I went with the cgroup-exporter suggestion without checking if telegraf supported it.
Looks like the exporter collects almost everything and doesn't currently support any config, telegraf doesn't collect anything by default, requires config.
There was a problem hiding this comment.
Does it make sense to collect everything? We mainly do alert monitoring rather than looking at performance metrics.
There was a problem hiding this comment.
I tried the plugin and some metrics had permission issues (telegraf user?) or were broken, exporter seems to work okay (or least doesn't seem to be missing anything notable).
Probably don't want to collect everything but might be easier to start with the exporter and try switching to the plugin later once we have a better idea of what we do want to collect rather than needing to configure it from the start.
There was a problem hiding this comment.
We mainly do alert monitoring rather than looking at performance metrics.
Maybe I should bring back the grafana dashboard? I removed it as no one was using it but maybe it would more useful now? IIRC people on matrix were talking about checking the load on the build boxes, dashboard might be a bit better than people running top.
zowoq
force-pushed
the
cgroups
branch
2 times, most recently
from
f3527c2
to
ec18abd
Compare
zowoq
force-pushed
the
cgroups
branch
3 times, most recently
from
ea26cdd
to
70b5371
Compare
No description provided.