Perlbal::Manual::FailOver − Fail Over mechanism in Perlbal
How Perlbal’s Fail Over mechanism works.
Whenever Perlbal fails to open a connection to a machine, it assumes that machine is not available and spreads the load between the remaining servers. It keeps trying to open connections to that server according to "error_retry_schedule".
If there are no available servers, Perlbal will keep trying to open connections to the servers until "idle_timeout" is reached, upon which it will close the connection to the user. Clients interpret that closing differently: most will assume a 500 HTTP code, a few assume a 200 OK with empty content.
For more information on parameters "idle_timeout" and "error_retry_schedule" see Perlbal::Manual::ReverseProxy.