Comment 18 for bug 1669468

Revision history for this message
Dr. Jens Harbott (j-harbott) wrote :

Not sure whether this is really the same issue still or if we should rather have a new bug for that, but the novnc config in tempest-multinode-full looks wrong to me.

https://logs.opendev.org/92/651492/7/check/tempest-multinode-full/ac24db2/controller/logs/etc/nova/nova-cpu_conf.txt.gz
my_ip = 10.211.2.5
[vnc]
server_proxyclient_address = 127.0.0.1
server_listen = 127.0.0.1
xvpvncproxy_base_url = http://10.211.2.5:6081/console
novncproxy_base_url = http://10.211.2.5:6080/vnc_auto.html

https://logs.opendev.org/92/651492/7/check/tempest-multinode-full/ac24db2/compute1/logs/etc/nova/nova-cpu_conf.txt.gz
my_ip = 10.211.2.46
[vnc]
server_proxyclient_address = 127.0.0.1
server_listen = 127.0.0.1
xvpvncproxy_base_url = http://10.211.2.5:6081/console
novncproxy_base_url = http://10.211.2.5:6080/vnc_auto.html

The section that generates this config in devstack however says:

        # Address on which instance vncservers will listen on compute hosts.
        # For multi-host, this should be the management ip of the compute host.

https://opendev.org/openstack/devstack/src/branch/master/lib/nova#L621

So I'm assuming the job only passes when the test VM gets scheduled on the controller and fails when it gets scheduled on compute1. For fixing the issue, the above config should contain the local IPs as set in my_ip instead of 127.0.0.1.