Hide Forgot
Description of problem: Even when DVR is used, central routers are still needed to handle some flows (e.g SNAT, IPv6). With RHOSP 10, DVR and L3 HA are not compatible. That means that routers are still going to be scheduled on the Controller nodes (and load shared between the L3 agents), but if one agent fails - all routers hosted by this agent fail as well. We want to make them work together to provide a better HA story.
Some documentation for the upcoming version: OSP 11 introduces a mix between distributed (DVR) routers and highly-available (HA) routers - these are DVR+HA routers. These routers can be treated as plain DVR routers as far as floating ips and east-west communication go - the only difference between DVR routers and DVR+HA routers are that the snat portion of DVR+HA routers is highly-available (using Keepalived, same as HA routers). In order to update a router to DVR+HA: 1. Turn the router off (neutron router-update <router_uuid> --admin-state-up=False 2. Update the router's attributes to indicate that it's DVR+HA (neutron router-update <router_uuid> --distributed=True --ha=True 3. Turn the router back on (neutron router-update <router_uuid> --admin-state-up=True In addition, routers can be created as DVR+HA to start with: * neutron router-create <name> --distributed=True --ha=True
Patch that allows migrations merged - 376550. Now pending on 383827 to create the DVR+HA job upstream, and a similar job in QECI, as well as patch 418276 that adds scenario tests for migration.
We can flip to ON_QA once we have an OSP11 build off 10.0.0.0b3.
Removing this one for RHOSP 11.