[OVS graceful restart] Tunnel flows are not saved after ovs agent restart
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Mirantis OpenStack |
Fix Released
|
Critical
|
Sergey Yudin |
Bug Description
On vxlan environment after restarting Open vSwitch agent on all controllers and computes connection to vm failed
Steps to reproduce:
1) Create 2 networks, router between them and boot 2 vms in different nets and on different computes
2) Start Ping between vms
3) Restart Open vSwitch agents on all controllers and computes with script:
ssh node-2 '. openrc && pcs resource disable p_neutron-
sleep 15
#Enable ovs-agents on controllers
ssh node-2 '. openrc && pcs resource enable p_neutron-
#Start ovs-agents on computes
fuel nodes | grep compute | awk '{print $1; }' | xargs -I@ ssh node-@ service neutron-
#Sleep $time - this value should be tuned manually to be sufficient for all agents going to "alive" state
sleep 5
Expected results: after restarting connection will not brake (1-3 ping packets loss is allowed)
Current result: connections brake, VMs can't renew ip addresses, can't ping gateway.
Root cause: tun-br lacks proper flows. During restart open vswitch agent we can see that flows appeared with 0x0 and then dissapeared
iso:
VERSION:
feature_groups:
- mirantis
production: "docker"
release: "7.0"
openstack_
api: "1.0"
build_number: "187"
build_id: "2015-08-
nailgun_sha: "4710801a2f4a6d
python-
fuel-agent_sha: "57145b1d880438
fuel-
astute_sha: "e24ca066bf6160
fuel-library_sha: "0062e69db17f8a
fuel-ostf_sha: "17786b86b78e5b
fuelmain_sha: "c9dad194e82a60
neutron vxlan, 3 controllers, 2 computes
Changed in mos: | |
status: | New → Confirmed |
tags: | added: on-verification |
Changed in mos: | |
assignee: | Eugene Nikanorov (enikanorov) → Sergey Yudin (tsipa740) |
Fix proposed to branch: openstack- ci/fuel- 7.0/2015. 1.0 /review. fuel-infra. org/10713
Change author: Eugene Nikanorov <email address hidden>
Review: https:/