Fuel version info (6.1 iso # 376): http://paste.openstack.org/show/214942/
Health checks fail on environment with Neutron+GRE (CentOS), because instances can't get IP address after boot:
http://jenkins-product.srt.mirantis.net:8080/view/6.1_swarm/job/6.1.system_test.centos.thread_7/116/testReport/junit/(root)/deploy_ceph_ha_nodegroups/deploy_ceph_ha_nodegroups/
http://paste.openstack.org/show/214945/
Steps to reproduce:
1. Create cluster with Neutron GRE and Ceph on CentOS
2. Add 3 controller + ceph nodes
3. Add 2 compute + ceph nodes
4. Select Fedora 3.10 kernel on 'Settings' tab
5. Deploy cluster
6. Run health checks (OSTF)
Expected result:
- environment passes health checks
Actual result:
- 2 checks fail ('Check network connectivity from instance via floating IP' and 'Launch instance with file injection') because after creating instance it's not reachable by internal/floating IP address
Here are errors I've found in neutron logs on controller (node-1):
http://paste.openstack.org/show/214940/
The same errors ('Failed to set-up gre tunnel port') present in neutron logs on all slave nodes. I tried to sniff traffic of management network for GRE packets, but tcpdump showed nothing. Also I was able to setup GRE tunnel between controller and compute manually and it worked fine, so I believe GRE isn't blocked by something. Ubuntu deployments aren't affected. Diagnostic snapshot is attached.
This issue affects all deployments with NeutronGre on CentOS:
http:// jenkins- product. srt.mirantis. net:8080/ job/6.1. system_ test.centos. thread_ 1/122/testRepor t/(root) /deploy_ neutron_ gre/deploy_ neutron_ gre/?