I didn't find anything on tracker, so I created this one: https://tracker.ceph.com/issues/72810 Zitat von Eugen Block <eblock@nde.ag>:
Hi *,
on a Squid (19.2.2) cluster with Grafana version 9.4.12 the "OSDs" panel on the "Ceph Cluster - Advanced" page shows some inconsistency, not sure if the attached screenshot will make it to the list.
We had to set an OSD out, waiting for the customer to make a decision about it. There are 126 OSDs in total, with one "out" the number of "in" OSDs should be 125, but it's still 126 although one OSD is shown as "out". I'm not going to comment on the colours (in are red, out is green).
Prometheus provides the correct status for the out OSD (0.0), but the expression is:
"alias": "In", ... "expr": "count(ceph_osd_in{cluster=~\"$cluster\", })",
which is just counting the entries, so it will always be the total number of OSDs.
I guess one could just adapt the expression for the "out" OSDs:
"alias": "Out", ... "expr": "sum(ceph_osd_in{cluster=~\"$cluster\", } == bool 1)",
I haven't checked tracker yet, is this already known? And when this is targeted, maybe fix the colours as well. To me it looks like it's still wrong in the latest version of ceph-mixin [0].
Regards, Eugen
[0] https://github.com/ceph/ceph/blob/346846543c6bfc93a360476c739580cd2344fec0/m...