Nova fails to live migrate instance with upper-case port MAC
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Compute (nova) |
Fix Released
|
Undecided
|
Unassigned | ||
neutron |
New
|
Medium
|
Unassigned |
Bug Description
Description
===========
When neutron port has MAC address defined in upper case and libvirt stores MAC in XML in lower case, migration is failed with KeyError:
```
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
9d85-6505-
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
Sep 30 10:31:38 cc-compute08-dx1 nova-compute[
```
Environment
===========
Ubuntu 20.04
Libvirt 6.0.0-0ubuntu8.14
Nova 22.2.3.dev2 (sha 4ce01d6c49f81b6
Neutron with OpenVSwitch
Changed in nova: | |
status: | New → In Progress |
adding neutron as i think neutron shoudl also be normaliasing the mac adress that users provide and alwasy storign it in lower case. a mac is technically a number not a string we just use hex encoding for human readablity so the caseing does not matter but it would be nice to at least consider moving this normalisation to the neutron api/db to avoid this problem.