Instance suddenly lost connectivity on tenant net, DVR with centralised snat
Affects | Status | Importance | Assigned to | Milestone | ||
---|---|---|---|---|---|---|
Mirantis OpenStack | Status tracked in 10.0.x | |||||
10.0.x |
Invalid
|
High
|
Kevin Benton | |||
7.0.x |
Fix Released
|
Critical
|
Sergii Rizvan | |||
8.0.x |
Invalid
|
High
|
Sergii Rizvan | |||
9.x |
Fix Released
|
High
|
Kevin Benton |
Bug Description
Detailed bug description: On a tenant network, instances lost connectivity to outside. In practice, they could ping the router internal interfaces, but not the external gateway anymore. In the flow, we were unable to trace exactly where the traffic got stuck.
Steps to reproduce: -
Reproducibility: Observed only once, so far
Expected results: VM not to lose connectivity
Actual result: VM unable to reach external networks
Workaround: Restart of ovs-agent with removal of namespaces didn't help
Impact: Instances lose connectivity
Description of the environment:
- Operating system: Ubuntu
- Versions of components: MOS 7.0
- Reference architecture: Neutron DVR with centralised snat
- Network model: VxLAN
- Related projects installed: Sahara, Ceilometer
Additional information: -
Upstream related issue in Neutron project:
https:/
Changed in mos: | |
importance: | Undecided → High |
tags: | added: support |
description: | updated |
description: | updated |
tags: | added: on-verification |
tags: | added: on-automation |
tags: | added: on-verification |
I have ovs-agent.log and l2-agent.log available upon request (Slack or mail)