Hello,
we're about to move 2 DNS LDAP backend (openldap-2.4.40) of sssd-1.12.4
from DNS round robin to a virtual ip on a load balancer. We don't use
SRV records.
So the plan is to go
from :
ldap_uri = ldaps://ldap.dom.ain
ldap.dom.ain having 2 IN A DNS records
to :
ldap_uri = ldaps://ldap.dom.ain
where ldap.dom.ain resolves in the load balancer hosted vip.
My understanding is that, in the target setup, if one of the backend is
shutdown or stop responding, sssd will still wait some timeout to
reconnect to the vip. So we still would have a time slice where
id <user>
would return no result (assuming <user> was not in the cache).
So the only gain for that matter would be that we would be garantied
that on the next backend connection, the backend would answer since the
loadbalancer would have pointed the connection to the working one.
Is there a way, other than tweaking the timout value, to avoid such
no-result hole ? (and without enmumarate = true either) ?
Thansk
--
Thomas HUMMEL