[Expired for neutron because there has been no activity for 60 days.]
** Changed in: neutron
Status: Incomplete => Expired
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to neutron.
Scalability of the legacy L3 reference implementation (OVS)
Status in neutron:
Here is the description and how this problem can be reproduced:
On a multi-node OpenStack instance with standard (legacy) l3 and no
- Create a new tenant
- Create a network/subnet, and a router for the tenant
- Connect the network to the router (with the external network set as the -
gateway as well)
- Create a VM on the network and assign a floating ip to it
- Ping the floating IP until the ping is successful
Time the above.
Repeat the above steps. As the number of tenants/networks/routers
increases (to several hundreds) so does the time it takes for these
operations to complete.
Time to create the tenant and the VM remains fairly flat.
The rest of the operations (setting up the networking) increases from
a few seconds at the beginning to more than a minute when the number
of routers grows to around a thousand.
Note that the test does not stress the messaging or API servers as
each iteration of above steps is executed after the previous iteration
has resulted in an operational router and floating IP.
The breakdown of various l3 operations (which seem to contribute to
the above) show the possible bugs reported here:
To manage notifications about this bug go to:
Mailing list: https://launchpad.net/~yahoo-eng-team
Post to : firstname.lastname@example.org
Unsubscribe : https://launchpad.net/~yahoo-eng-team
More help : https://help.launchpad.net/ListHelp