On 22 January 2013 17:31, Zachary Stern <[email protected]> wrote: > > The params are listed in the HAproxy manual. > > Boris, my params are not a magical fix to make haproxy instantly detect > downtime. As you can see, my interval is set to 30 seconds. > > Please read the manual to understand how those parameters function. It's > not obvious. > > http://cbonte.github.com/haproxy-dconv/configuration-1.4.html#5-inter
Also perhaps take a look at the layer4 version of "observe": http://cbonte.github.com/haproxy-dconv/configuration-1.4.html#observe (I don't suggest using the layer7 version as I believe it allows an attacker who discovers a URI that produces a 500 to fail your backends for you.) "observe layer4" mitigates server failure when you're using a non-insane check interval. Though, as a data point, I insist my devs write status endpoints for HAProxy's use that can withstand 1Hz checks (i.e. 2 active/passive HAProxy nodes @ 2 second intervals). 30 seconds isn't frequent enough, for my money. YMMV. Jonathan -- Jonathan Matthews // Oxford, London, UK http://www.jpluscplusm.com/contact.html
