[DVR] Not correct rescheduling while agents are unstable
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Mirantis OpenStack |
Fix Released
|
High
|
Oleg Bondarev | ||
7.0.x |
Won't Fix
|
Medium
|
Sergii Rizvan | ||
8.0.x |
Fix Released
|
High
|
Oleg Bondarev |
Bug Description
After reseting primary controller connection to the internet and other vms was lost. This case didn't appeared permanently
Steps:
1) Create net, subnet and connect them to external net with DVR router
2) Launch vm in new net and associate floating ip
3) Check connection to internet
4) Reset primary controller
5) Wait some minutes
6) Check connectivity
Expected result: ping 8.8.8.8 works properly
Current result: Message "Network is unteachable"
iso (vxlan + l2 + dvr):
{"build_id": "296", "build_number": "296", "release_versions": {"2015.1.0-7.0": {"VERSION": {"build_id": "296", "build_number": "296", "api": "1.0", "fuel-library_sha": "f2eef7717b15c6
Changed in mos: | |
status: | New → Confirmed |
The bug is that if only 'dvr' (on computes) agents are active at the moment of DVR router rescheduling - the router will be rescheduled only on them and rescheduling will be considered as successful. The fix would be to fail if snat was not scheduled.