@gnaponie Please take a look.....
/cc @jkaluza

@fivaldi Can you please write down some background or context of this monitor API? I think that could be helpful to understand the implementation.

I think I like how this is integrated, but I'm not sure what to do with prometheus_client dependency. So far we are installing only RPM dependencies in Freshmaker's container image and prometheus_client is not installed. I think I'm not against adding "pip install prometheus_client" there for Freshmaker, but for ODCS or MBS (which does not run in container image), it would be great to have package with prometheus_client.

Did you check if this is working properly? Because I tried to do the same, but it was every time creating a new temp directory and making mess.. so I just chose "/tmp". I'm not proposing to change this, I would just like to know if it is working properly for you, so maybe we can change it also in Greenwave.

I like it. +1
But there's no generic "metrics" endpoint right? You created specific endpoints for each type.

FYI - python prometheus client is packaged: https://koji.fedoraproject.org/koji/packageinfo?packageID=27137

Hi.

I changed the endpoint to /monitor/metrics. We can use /monitor/<...> so that we have the metrics and we can re-use the same API (soon) for other stuff, like logs, tracebacks, self-tests etc. as proposed in the Monitoring doc. As a benefit/side effect, we'll have it all under one hood.

@gnaponie The /tmp/.... dir (if set via the env var), should be the same for all workers.

it looks good to me

Commit 5a9932c5 fixes this pull-request