<div dir="ltr"><div class="gmail_extra"><div class="gmail_quote">On 6 March 2014 14:54, Alvise Dorigo <span dir="ltr"><<a href="mailto:alvise.dorigo@pd.infn.it" target="_blank">alvise.dorigo@pd.infn.it</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div style="word-wrap:break-word"><div>As stated in the mentioned thread there's no solution for making L3 agent HA with the active/active paradigm. But I have to use the active/active paradigm.</div>
</div></blockquote><div><br></div><div>There is a blueprint for a fully active-active L3 Agent that's close to implementation:</div><div><a href="https://blueprints.launchpad.net/neutron/+spec/l3-high-availability">https://blueprints.launchpad.net/neutron/+spec/l3-high-availability</a><br>
</div><div><br></div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div style="word-wrap:break-word">
<div>In the mentioned thread, Jay suggests to use a tool to replicate l3 agent's status to the other node upon failure of the current node (I haven't used this tool yet).<br></div></div></blockquote><div><br></div>
<div>We use this in production and it works well. Essentially it doesn't give you true active-active high availability, but it does give you automated redundancy and you are able to use multiple L3 Agents actively.</div>
<div>The path for packets will only every use one L3 Agent at a time, but if an Agent goes down, the tool will migrate the L3 Agent servicing the virtual routers to an alternative active agent. The tenant/project's L3 traffic will be lost only for the duration of the down-time detection and move. If the tool is run every minute in cron, the detection will happen within a minute. It has a random wait period of less than a minute before starting a move. The time it take for the move itself to complete depends on how many virtual routers need to be moved.</div>
<div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div style="word-wrap:break-word"><div></div><div>
Reading the HA guide by Openstack (<a href="http://docs.openstack.org/high-availability-guide/content/_running_neutron_metadata_agent.html" target="_blank">http://docs.openstack.org/high-availability-guide/content/_running_neutron_metadata_agent.html</a>) there's not a solution active/active-based also for the metadata agent.<br>
</div></div></blockquote><div><br></div><div>The metadata agent is stateless and can be implemented in an active-active manner with load balancing using something like haproxy. We do this successfully in production.</div>
</div></div></div>