Hi guys,
I'm looking for feedback about OpenDayLight and OpenStack at scale.
We are running several OpenStack (Rocky) platforms, some of them have
thousands of instances and so thousands of Neutron ports, networks,
subnets, etc...
Neutron maintenance are very painful for us because it doesn't really
work well at this scale.
To give metrics, restarting Neutron server will cause Neutron Open
vSwicth and DHCP agents to sync (ports status, etc...) during 10 hours!!
During these 10 hours, instance creation fails because Neutron agents
are not UP which make our customers not really happy :)
Our setup is pretty standard:
- DVR to avoid bottleneck on control nodes
- Open vSwitch 2.11
- Using "vsctl" instead of "native" driver (we had scale issue in the
past)
- Using rootwrap-daemon to avoid some bottlenecks
- Timeouts have been increased
Basically we are looking for alternatives and OpenDayLight is one of
them, Red Hat integrated it into OSP since Pike and the integration
within OpenStack improved a lot since the past few months.
We are looking for feedback from people using OpenDayLight with
OpenStack at scale:
- How does it handle the load ?
- How the maintenance are painful ?
- Does it scale easily ?
- How the upgrade process is ?
Thanks :)
Gaëtan
_______________________________________________
Discuss mailing list
[email protected]
https://lists.opendaylight.org/mailman/listinfo/discuss
