Bug #47537
closed
Prometheus rbd metrics absent by default
Added by Daniël Vos over 3 years ago.
Updated 4 days ago.
Category:
prometheus module
Description
Hello,
According to https://docs.ceph.com/en/latest/mgr/prometheus/#rbd-io-statistics statistics should be collected for all rbd enabled pools by default
My cluster has a osd pool by the name of "LXD" that contains rbd images. `ceph config get mgr mgr/prometheus/rbd_stats_pools` is empty and according to the docs this means that it should collect/export metrics for all pools but unfortunately it does not.
When I explicitly tell it to export metrics for the pool 'LXD' as such: `ceph config set mgr mgr/prometheus/rbd_stats_pools 'LXD'` ceph_rbd_* metrics pop up in my prometheus.
And ofcourse when I empty the config option with `ceph config rm mgr mgr/prometheus/rbd_stats_pools` my rbd metrics disappear
Are the docs incorrect or is this a logic error somewhere? :)
ceph config get mgr mgr/prometheus/rbd_stats_pools
LXD
ceph config rm mgr mgr/prometheus/rbd_stats_pools
<metrics disappear>
ceph config set mgr mgr/prometheus/rbd_stats_pools 'LXD'
<metrics appear again>
ceph osd pool application get LXD
{
"rbd": {}
}
Running on Ceph 15.2.4
- Category set to prometheus module
Actually, the doc strings in the module says:
If no namespace is specifed the stats are collected for every namespace in the pool. The wildcard '*' can be used to indicate all pools or namespaces
Daniël, can you try to set wildcard value like this?
ceph config set mgr mgr/prometheus/rbd_stats_pools '*'
- Status changed from New to In Progress
- Target version set to v17.0.0
- Source set to Community (user)
- Backport set to pacific octopus
- Severity changed from 3 - minor to 4 - irritation
- Pull request ID set to 43216
- Affected Versions v14.2.22, v15.2.14, v16.2.6 added
- Affected Versions deleted (
v15.2.4)
Konstantin Shalygin wrote:
Actually, the doc strings in the module says:
If no namespace is specifed the stats are collected for every namespace in the pool. The wildcard '*' can be used to indicate all pools or namespaces
Daniël, can you try to set wildcard value like this?
[...]
Thank you! This option would be a nice one to add to the docs on docs.ceph.com. Would there be a performance impact if the wildcard were to be made the default?
I see now you already made the addition to the docs. Thanks!
By default is better to be off, because on big clusters it's may be performance impact or just be a overhead (e.g. just unneeded info)
When somebody verify PR - record will be landed to docs
- Status changed from In Progress to Resolved
- Assignee set to Konstantin Shalygin
- % Done changed from 0 to 100
Also available in: Atom
PDF