Closed jgruber closed 7 years ago
Appropriate configuration working in v9.3.1b, however can still reproduce stranded node addresses, SNAT pools / translation addresses, self-IPs, FDB records, route domains, and partitions.
Closing this one. Will open a separate issue for the config stranding.
Title: Creation of 10 consecutive pools with 2 members each on a BIG-IP leads to failures.
Attachments: None - 100% reproducible with various log entries per run. I will post a clean log on Monday. Details: Creation of 10 consecutive pools of 2 members each on a BIG-IP pair results in the creation of 10 pools with 1 and sometime 2 members. This creates a difference in the LBaaS model in Neutron verse the BIG-IP Configuration. Deletion of the pool members leads to stranded Node Addresses and a tenant partition which does not have any LBaaS resources assigned in Neutron.
OpenStack Release
9.3.0
Description
Creation of 10 consecutive pools of 2 members each on a BIG-IP pair results in the creation of 10 pools with 1 and sometime 2 members. This creates a difference in the LBaaS model in Neutron verse the BIG-IP Configuration. Deletion of the pool members leads to stranded Node Addresses and a tenant partition which does not have any LBaaS resources assigned in Neutron.
Deployment
Deploy LBaaS with a pair of BIG-IPs.
Use the following Heat template to orchestrate the 10 pool deployment, with input parameters adjusted to your cloud.
http://repo.mydemo.rocks/templates/labs/deploy_lbaas_20_web_servers_10_pools.yaml
You will need a web server glance image, and 2 subnets defined (one for the load balancer) and another for all nova server deployments.
Monitor the BIG-IP for pool members for difference between Neutron and BIG-IP once the neutron agent for the BIG-IP has reached a request depth of zero (thinks it is done provisioning).
Delete the Heat stack and observe stranded Node Addresses on the BIG-IP.
100% reproducible.
Agent Version
9.3.0
Operating System
Centos7
OpenStack Release
Mitaka