I think you have a problem similar to one I have. The priority of beacons seems
very low. As soon as something gets busy, beacons are ignored or not sent. This
was part of your log messages from the MDS. It stopped reporting to the MONs
due to laggy connection. This laggyness is a result of
Hi Francois,
there is actually one more parameter you might consider changing in case the
MDS gets kicked out again. For a system under such high memory pressure, the
value for the kernel parameter vm.min_free_kbytes might need adjusting. You can
check the current value with
sysctl
I can vaguely remember having an argument with zabbix developers on how
inefficient they were putting metrics in the mysql database, causing
unnecessary high load. I guess this was the time of mysql 3. I never can
let go the feeling that if people are applying a type of logics, they
will
Thanks for the tip,
I will try that. For now vm.min_free_kbytes = 90112
Indeed, yesterday after your last mail I set mds_beacon_grace to 240.0
but this didn't change anything...
-27> 2020-06-06 06:15:07.373 7f83e3626700 1
mds.beacon.lpnceph-mds04.in2p3.fr MDS connection to Monitors appears
Trying to configure Zabbix module in Octopus 15.2.3.
CentOS 8.1 environment. Installed zabbix40-agent for CentOS 8.1 (from epel
repository). This will also install zabbix_sender.
After enabling the Zabbix module in Ceph, I configured my Zabbix host and
Zabbix identifier.
# ceph zabbix config-set