MTU mismatch between o-hm0 and lb-mgmt-net
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Charm Guide |
Triaged
|
Undecided
|
Unassigned | ||
OpenStack Octavia Charm |
Fix Released
|
High
|
Edward Hope-Morley | ||
2023.1 |
Fix Released
|
Undecided
|
Unassigned | ||
Yoga |
Fix Released
|
Undecided
|
Unassigned | ||
Zed |
Fix Released
|
Undecided
|
Unassigned |
Bug Description
lb-mgmt-net is created by the charm with mtu 8942 while on o-hm0 charm sets mtu 1500. This is causing heartbeats of certain LBs to be dropped forcing octavia to fail over them.
This seem related to https:/
Some LBs, because of the number of members/
i.e.
12:06:56.706585 fa:16:3e:56:a6:05 > fa:16:3e:cb:ba:9e, ethertype IPv6 (0x86dd), length 1596: fc00:51be:
See 1534 > 1500 and o-hm0 never receives them.
We worked around that running on all octavia:
ovs-vsctl set interface o-hm0 mtu_request=8942
Changed in charm-octavia: | |
status: | Fix Committed → Fix Released |
I had the same situation when the lb-mgmt-net was set to use 9042 MTU and o-hm0 is set to use 1500 MTU.
The Octavia units will show the drop packets:
[2065317.692684] o-hm0: dropped over-mtu packet: 1744 > 1500
[2065339.573539] o-hm0: dropped over-mtu packet: 1654 > 1500
[2065347.747633] o-hm0: dropped over-mtu packet: 1744 > 1500
[2065375.092206] o-hm0: dropped over-mtu packet: 1653 > 1500
[2065405.176370] o-hm0: dropped over-mtu packet: 1655 > 1500
If the lb-mgmt-net is set to use the same MTU as o-hm0, for example, 1500 it will work using defragment.
During the problem, Octavia Health starts the failover of the amphoras after 60 seconds.
In this case, the LB has more than 60 members.
[Workaround]
The workaround described on this LP works, but the final solution would be the charm check the MTU config lb-mgmt-net and set the same config for o-hm0 interface.