Yes, lb group provisions haproxy (not the openshift router) to do TCP load balancing between external clients and the api servers on openshift_master_api_port which defaults to 8443.
On Tue, Oct 4, 2016 at 9:03 AM, Andrew Lau <[email protected]> wrote: > Hey Scott, > > Thanks for the in depth insight - really appreciate it. > > Not using the loadbalancer/haproxy from openshift-ansible. Is it just port > 8443 which the internal haproxy exposes? > > Thanks, > Andrew > > On Tue, 4 Oct 2016 at 23:32 Scott Dodson <[email protected]> wrote: >> >> Andrew, >> >> Our ops teams remove the endpoint from the kubernetes service and then >> restart the master. When the master starts again it automatically adds >> its api endpoint back into the kubernetes service. This process would >> proactively remove the endpoint but if that weren't done the endpoint >> leasing introduced in 1.3 should automatically remove the endpoint >> after a brief period of time. They also manually manage their load >> balancer to remove the api server there when rolling a given master, >> we don't have anything like that for the load balancer provisioned >> using openshift-ansible, I don't know if you're using that load >> balancer or not. >> >> -- >> Scott >> >> On Sat, Oct 1, 2016 at 2:42 AM, Andrew Lau <[email protected]> wrote: >> > Is there something like node evacuate for master hosts >> > >> > If we want to restart a master host in a HA cluster, for whatever >> > reason, >> > it'll cause some temporary failures like pod/service DNS lookup. >> > >> > With routers it's been possible to remove it from the external DNS >> > roundrobin or LB before performing the restart. >> > >> > _______________________________________________ >> > users mailing list >> > [email protected] >> > http://lists.openshift.redhat.com/openshiftmm/listinfo/users >> > _______________________________________________ users mailing list [email protected] http://lists.openshift.redhat.com/openshiftmm/listinfo/users
