[RFE] Set lost port status to down during neutorn ovs agent restart
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
oslo.messaging |
Expired
|
Undecided
|
Unassigned |
Bug Description
Port status will be reset to DOWN if ovs port is removed from br-int bridge while neutron-ovs-agent is running. On the contrary, if the ovs port is removed before neutron-ovs-agent started, then ovsdb monitor will not notify ovs-agent and thus ovs-agent will not flag this port to be DOWN.
I am bringing this up because we have a use case where we need to deal with situation all nodes in cluster might be cut off power at same time. So during node reboot, nova-compute will not be coordinated with ovs-agent and dhcp-agent, which means vm could lost its IP address.
One workaround I am thinking of is to unplug port (not through neutron api, just remove port), wait for port status to be down, and then replug the port again. The idea is neutron-ovs-agent should flag missing port as DOWN and reset its to ACTIVE if it is recreated again.
Changed in oslo.messaging: | |
status: | New → Incomplete |
summary: |
- Set lost port status to down during neutorn ovs agent restart + [RFE] Set lost port status to down during neutorn ovs agent restart |
[Expired for oslo.messaging because there has been no activity for 60 days.]