Backend health right after launch

Martin Goldman me at
Fri May 29 19:03:41 CEST 2009

Hello all,

I'm trying varnish for the first time, and I've noticed that for the first
several minutes after starting the service, every request 503s out. After
that, requests start working. I'm guessing this is because I'm doing backend
polling with a config like this:

backend web1 {
   .host = "server1";
   .port = "80";
   .probe = {
      .url = "/ping.php";
      .interval = 1 m;
      .timeout = 2 s;
*      .window = 5;
      .threshold = 3;
*   }

With this config, a backend isn't considered healthy unless it passes the
health check 3 times. Since I'm only doing the check once per minute, it
takes 3 minutes for the backends to become healthy.

I think the "3 out of 5" check is a good one to have during normal server
operations, but I'd prefer if it didn't apply right after startup, since the
service is then guaranteed to be unavailable for several minutes despite
probably being healthy. I understand why it works this way, but I'm just
curious, is there a way around this?

Many thanks,
Martin Goldman
-------------- next part --------------
An HTML attachment was scrubbed...

More information about the varnish-dist mailing list