On Sat, Apr 3, 2010 at 9:44 AM, Rainer Jung <rainer.j...@kippdata.de> wrote: > On 31.03.2010 23:33, Mohit Anchlia wrote: >> >> On Wed, Mar 31, 2010 at 2:27 PM, Rainer Jung<rainer.j...@kippdata.de> >> wrote: >>> >>> Which version? >>> >>> If 1.2.28 or newer, look for error_escalation_time in >> >> Actually I am on 1.2.27 so can't use error escalation time. Is there >> any other way I can do it? I am seeing real odd behaviour with mod_jk. > > No, there was a change in 1.2.27: > > LB: Do not put loadbalancer node in error state if there is opened channel. > This fixes the bug when new connection fails due to busyness, causing opened > connections fail stickyness. This brings back per-node busy counter and > private state array for each request. We can mark the state as error for > failover to work while still operating and reporting node as OK if there are > opened working connections.
Thanks! this is what I am not understanding that when box is down there are no open connections to that box, because box is down. So why will that worker not go in ERR state. And err 111 and 115 clearly indicate that. Other question is, as soon as box comes backup it goes in ERR state as long as httpd2 is down. So it seems odd that in first case where it really should put worker in ERR state, it doesn't do that. > > That change motivated the addition of error_escalation_time in order to be > able to stick with the previous behaviour. You will have to update to 1.2.28 > or beyond. > > Regards, > > Rainer > > --------------------------------------------------------------------- > To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org > For additional commands, e-mail: users-h...@tomcat.apache.org > > --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org For additional commands, e-mail: users-h...@tomcat.apache.org
