Hi,
we are currently setting up our keycloak cluster (version 1.9.1) for the production
environment. The cluster consists of two keycloak nodes running in standalone-ha mode
whereby jgroups is configured to use tcping instead of multicast and loadbalancing is done
via modcluster. So far so good, the responses of the KC rest services are fast enough and
everthing works as expected.
The only problem what we still have is that the admin console is very slow if it is
invoked on the cluster ip. For example the request to render the realm overview page
takes more than 5 seconds
(https://${CLUSTER}/auth/admin/realms<https://$%7bCLUSTER%7d/auth/admin/realms>).
The problem is additionally increased because the GET realms request is sent up to four
times.
If we take out one keycloak node in the loadbalancer the admin console is essentially
faster. Can anyone help or does anybody have an idea what can be the misconfiguration
here? Or is this a known issue that the admin console is slow in cluster mode?
Thanks in advance!
Cheers,
sascha