We are testing Openstack LBaaS plugin to manage F5 VE/Hardware LTM .
Customer is using Centos kilo version openstack and installed F5 LBaaS v1.0.10 plugin. By now , we had tested multi agent support for F5 LBaaS plugin , here is the status:
1> 1 controller with 1 LBaaS agent to manage 1 LTM
This is ok
2> 1 controller with 1 LBaaS agent to manage 1 pair HA LTM
This is not work , we can adding pools and succeed in deliver to LTM , but when we try to adding vip , the system will automatic delete the whole partition on LTM after deliver to LTM .
3> 2 controller each had 1 LBaaS agent to manage 1 LTM
This is also not work , we can adding pools but failed to adding VIP , same result , VIP was deliver to LTM once , and after 10-30s , LTM will delete the VIP.
We had try to enable evironment_prefix == uuid , still get the same issue.
And also , for step3 , we double confirmed if enable each of the controller only , everything is OK.
So , I just wonder , have anyone here test multi agent solution for Openstack LBaaS ?
If I understand you correctly, scenario 2 uses a single LBaaS agent to control a cluster of two BIG-IP LTMs. This is supported and tested. Standalone, HA-pair and HA-scalen operations are exercised in regression testing performed within the engineering team. I see that you have 'f5_ha_type = standalone' in your .ini file, which controls how the agent interacts with the BIG-IP(s). You'll need to set 'f5_ha_type = pair' and restart the agent. Further details on HA configuration can be found in f5-oslbaasv1-readme.pdf, which should be included in the F5 LBaaS image bundle.
Scenario 3 looks like the same bug reported to the product team last week. The team is currently working on both a fix and method to reproduce internally to verify the fix. We're not aware of a workaround at the moment that supports the use of multiple controllers and agents.