Yu Wei <yu20...@hotmail.com> writes: > F0924 22:58:50.301065 1 start_api.go:68] dial tcp 172.16.211.60:2379: > connect: connection refused
the api tries to contact etcd on 172.16.211.60:2379 but > 2019-09-24 20:34:26.273020 I | embed: serving client requests on > 9.30.119.88:2379 but etcd listens for client request on 9.30.119.88. a wild guess is that your hostname resolves to 172.16.x.x but your default route points to 9.30.x.x. for a longer explanation why i think this is the case read on :-) the listing address for etcd is configure in /etc/etcd/etcd.conf (ETCD_LISTEN_CLIENT_URLS). etcd.conf gets created by the openshift-ansible role etcd (/usr/share/ansible/openshift-ansible/roles/etcd/templates/etcd.conf.j2), it uses etcd_listen_client_urls. etcd_listen_client_urls uses etcd_ip which gets set via openshift.common.ip, that's a custom fact to be found in openshift_facts.py (.../roles/openshift_facts/library/openshift_facts.py). in get_defaults() ip_addr gets set to the standard ansible fact ansible_default_ipv4.address, this is then used common.ip. so it seems that ansible thinks that 9.3.30.119.88 is your default interface, i'm not 100% sure, but this might be because your default route points to this nic (ip route sh). the config option for the api to contact etcd is in /etc/origin/master/master-config.yaml, etcdClientInfo.urls. openshift-ansible uses openshift_master_etcd_hosts for this, afaik this is set via the custom fact openshift.common.hostname (once again in openshift_facts.py) for figuring out the public hostname (it reuses ansible_nodename and ansible_fqdn). hopefully this is not too confusing... hth toni _______________________________________________ users mailing list users@lists.openshift.redhat.com http://lists.openshift.redhat.com/openshiftmm/listinfo/users