Thanks for the tips, I will evaluate the Amazon ELB offering since we are using juju-deployed mongrel2 on EC2 now.
-- Paul Eipper On Sun, May 10, 2015 at 1:12 PM, Henry Finucane <[email protected]> wrote: > Howdy, > In general, anything for doing load balancing between layer 7 proxies, like > haproxy, should be applicable to mongrel2. > On May 8, 2015 7:31 AM, "Paul Eipper" <[email protected]> wrote: >> >> Greetings, >> >> Is there a recommended/common pattern for deploying mongrel2 and >> enabling high availability and increasing the reliability of the >> service, that is, reduce the single point of failure of having one >> mongrel2 entry point machine? > > Keepalived is popular for this. If you're on Amazon's cloud, ELB is the way > to go. > >> Is it possible to have redundant mongrel2 machines, in a way that if >> one goes down, the other can reliably continue/respond the connections >> from the previous? > > I don't think so- this is a pretty hard problem, especially if you value > responsiveness. > >> My understanding is that a mongrel2 instance stores connection state, >> so I am not sure if this is possible? >> >> Should the client and backend solve this by resending on timeout and >> being idempotent? > > Yes =) > >> Ideas regarding this are welcome. >> >> att, >> -- >> Paul Eipper
