Used Build : 2.10-39~havana on Ubuntu 12.04
The roledef in testbed.py looked like below :
#Role definition of the hosts.
env.roledefs = {
'all': [host1, host2, host3, host4, host5],
'cfgm': [host1, host2, host3],
'openstack': [host1, host2, host3],
'control': [host1, host2, host3],
'compute': [host4, host5],
'collector': [host1, host2, host3],
'webui': [host1, host2, host3],
'database': [host1, host2, host3],
'build': [host_build],
#'storage-master': [host1],
#'storage-compute': [host4, host5, host6, host7, host8, host9, host10],
# 'vgw': [host4, host5], # Optional, Only to enable VGW. Only compute can support vgw
# 'tsn': [host10], # Optional, Only to enable TSN. Only compute can support TSN
# 'toragent': [host10], Optional, Only to enable Tor Agent. Only compute can
# support Tor Agent
# 'backup':[backup_node], # only if the backup_node is defined
}
host1 = 10.204.74.1
host2 = 10.204.74.2
host3 = 10.204.74.3
host4 = 10.204.74.4
host5 = 10.204.74.5
Post installation, I could see vRouter agent had established XMPP connections with host2 and host3. Please see below :
Host1:
root@budweiser:/opt/contrail/utils# netstat -an | grep :5269
tcp 0 0 0.0.0.0:5269 0.0.0.0:* LISTEN
Host2:
root@corona:~# netstat -an | grep :5269
tcp 0 0 0.0.0.0:5269 0.0.0.0:* LISTEN
tcp 0 0 10.204.74.2:5269 10.204.74.5:35461 ESTABLISHED
tcp 0 0 10.204.74.2:5269 10.204.74.4:41753 ESTABLISHED
Host3:
root@heineken:~# netstat -an | grep :5269
tcp 0 0 0.0.0.0:5269 0.0.0.0:* LISTEN
tcp 0 0 10.204.74.3:5269 10.204.74.5:42164 ESTABLISHED
tcp 0 0 10.204.74.3:5269 10.204.74.4:52994 ESTABLISHED
Host2 and Host3 were having connections with Host4 and Host5.
I plugged Host2 out of the network. Within 2-3 minutes, I could see that Host1 and Host3 established XMPP connections with Host4 and Host5. This is fine.
However, the dashboard (Contrail UI) looks unstable with red dots pointing "system processes unavailable". At times, under Contol Nodes - it showed Host2 and Host3 even though Host 2 was out of the network. The dashboard did not stabilize at all.
I went ahead and rebooted the nodes to fix the issue of dashboard. I feel from the functionality perspective there is no issue as the vRouter agents correctly established connections with Host1 when Host2 was pulled out from the network. However, dashboard reports incorrect status. It is easily reproducible in the lab.
Please investigate.
Can you please update in which version of Contrail will this bug be addressed ?
- Sandeep.