- linking to prometheus (http://hostname:9095/graph?...),for example
from Monitoring -> URL field (or the alerts' link) uses single
hostname _and_ http (which may be configured to be https, I haven't
checked yet). Short hostname doesn't work and http doesn't either due
to HSTS on the main domain.
You can set the prometheus URL with the dashboard command I mentioned
and specify the FQDN. If your grafana/prometheus host has changed then
the URL has to be updated, of course. I'm not sure if you want to
failover these services, the database would be lost, wouldn't it?
Zitat von grin <cephlist@xxxxxxxxxxxx>:
On Fri, 22 Apr 2022 06:54:33 +0000
Eugen Block <eblock@xxxxxx> wrote:
Hi,
> They are either static (so when the manager moves they become dead)
> or dynamic (so they will be overwritten the moment the mgr moves),
> aren't they?
there might be a misunderstanding but the MGR failover will just
redirect your dashboard access to the new active MGR. You can set
that option like this (redirect is default):
$ ceph config set mgr mgr/dashboard/standby_behaviour "redirect"
But a failover won't overwrite any URLs like grafana or prometheus,
the only URL that changes is the dashboard URL. Usually, you have
only one alertmanager, prometheus and grafana server for your cluster
so a static URL works fine.
I'm really not sure if we're on the same page here, if not please
clarify.
You are probably right that there are multiple cases here.
- mgr redirect: it redirects to the wrong hostname (I believe I have
seen bare hostname before but now it seems to use the reverse dns,
which isn't what I wanted [ugly] but it works). (A possible problem
here is that the host may have multiple IP and orch picks the wrong
one (the subnet between the server and the containers is separate
from the one used to access the web on them), but that was not
strictly my problem now.)
I think it does not use the /etc/hosts name.
- grafana URL, which includes the graphs in dashboard. I am not sure
whether it's dynamic or not, but it definitely have been updated by
"something" to the bare hostname in the past. Maybe it was not the
failover, I am not sure that change is logged anywhere.
- linking to prometheus (http://hostname:9095/graph?...),for example
from Monitoring -> URL field (or the alerts' link) uses single
hostname _and_ http (which may be configured to be https, I haven't
checked yet). Short hostname doesn't work and http doesn't either due
to HSTS on the main domain.
Right now I have multiple managers and one prometheus, alertmanager and
grafana (but they may have been failed over to different hosts due to
server reboots).
Peter
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
