HA: OpenStack cluster became non-functional when management interface of active controller has incorrect configuration
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Fuel for OpenStack |
Confirmed
|
Critical
|
Unassigned |
Bug Description
Steps To Reproduce:
1. Deploy OpenStack cloud in HA mode using Fuel 5.0.1 or Fuel 5.1 test ISO (CentOS, 3 controllers + mongo, 2 computes, Nova Network with DHCP, Install Ceilometer and Sahara).
2. Run OSTF tests and verify that all works fine.
3. Login to active controller and re-configure management interface IP address, for example:
eth0.101 Link encap:Ethernet HWaddr 64:C5:1E:B8:2A:0A
inet addr:192.168.0.2 Bcast:192.168.0.255 Mask:255.255.255.0
inet6 addr: fe80::66c5:
eth0.101-hapr Link encap:Ethernet HWaddr CE:5E:96:0F:B7:2D
inet addr:192.168.0.2 Bcast:0.0.0.0 Mask:255.255.255.0
inet6 addr: fe80::cc5e:
_______
and run the following command to change IP address:
ifconfig eth0.101 192.168.0.100
4. Start OSTF tests again.
Observed Result:
All OSTF tests will fail, because HA cluster attempts to connect to the first controller via management interface. The failover scenario doesn't work for this case.