Ceph dashboard performance data (Prometheus etc)

[ozdjh]

Hi

Installing the dashboard on our lab cluster was very simple. But it doesn't show any of the performance or monitoring data I was looking for. The Ceph documentation says to just enable it with cephadm but that's obviously a bad idea on a PVE orchestrated ceph cluster.

So is there a PVE friendly way to install and configure the Prometheus and Grafana components that'd be setup automatically if deploying via cephadm?

 

[Ting]

did you resolve your issue? I do not see anybody response.

Wondering why you said it was a bad idea to enable the dashboard under pve? I have an issue, every time I enable dashboard, all my ceph manager daemon crashs, and can not be restarted, no one response to your post.

 

[ozdjh]

No, I didn't resolve this with the dashboard. We ended up pulling metrics out of the ceph command line tools and ingesting them into our monitoring system so we can have reasonable visibility and alerting.

The "bad idea" isn't about enabling the dashboard, it's about using cephadm to enable the dashboard. PVE has it's own orchestration for Ceph provisioning, so running cephadm to do anything has the potential of messing up the config files that pveceph has created.

 

[Ting]

thanks for your quick reply.

 

[godfather007]

So if i get it right, using ceph-dashboard, how tempting it would seem, has the ability to break the existing ceph-environment?

So let's say, i want to use the same hardware for storing S3 data through RGW, or using NFS-ganesha, i better do this manually?

 

[ozdjh]

No, you are missing the point. Using cephadm to install any ceph add-ons "has the ability to break the existing ceph-environment". You cannot use 2 different orchestration tools for the same ceph cluster.

 

[godfather007]

So it is not possible (or dangerous) to manually add an S3 pool for app-X? Even without the official tools as Grin has written in the wiki?

 

[ozdjh]

I'm not sure why you aren't understanding this. You can manually do whatever you want. Just don't use another orchestration tool like cephadm to install other features or add-ons as it may overwrite the configuration files generated by pveceph.

 

[godfather007]

Love your friendly style, thanks that you're warning me against using the other orchestration tool and stick to the basics.

 

[alynio]

try use https://docs.ceph.com/en/octopus/cephadm/monitoring/
but you need setup prometheus and grafana servers - I'm using lxc images

 

[kellogs]

is there anyway to enable the prometheus plugin on ceph using pveceph command?

 

[gregmans]

is there anyway to enable the prometheus plugin on ceph using pveceph command?

You can do this with the regular ceph command.

 ceph mgr module enable prometheus

 

[kellogs]

You can do this with the regular ceph command.

 ceph mgr module enable prometheus

I am worried it might break “pvecepgh” since the setup was done using proxmox and not cephadm?