Argh. Sorry, obviously, I got the configuration statement wrong :) server ... inter 5s fall 3... would be the way to set it up so it checks every 5 seconds and after 3 failures, haproxy marks that particular server as down.
But still, during that 15s, what happens to incoming connections? There's a reference to fastinter, which I'm guessing is the amount of time in between checks after the first failure (thus transitionally down) but before the third failure (finally marked as down for real). At least, I'm guessing that's what that timeframe is for. So I suppose I could narrow that 15s window down a bit. After the first failure, I could use fastinter to maybe 2s. So the total time would be 5+2+2 = 9s rather than 15s. The downinter I tend to set fairly high. Since I don't need it checking a dead system every 5s. It's probably dead for a reason :) My apologies for the confusion, and thank you in advance for any clarification, PH On Tue, Dec 22, 2009 at 2:27 PM, Paul Hirose <[email protected]> wrote: > Using the server ... inter 5s fail 15s for example, if I understand > it, will do a health check every 5 seconds. If it fails three times > in a row (down for 15s or more), then haproxy concludes the realserver > as dead. > > What happens to incoming connections during that 15s? Do they get > queued up and eventually go to the other server(s) in that pool? Do > they just continue to try connecting to the (possibly) dead server? > Or do they just eventually timeout and return whatever timeout-related > error code to the originating client? > > Thank you, > PH >
