[Varnish] #837: varnishadm purge.list fails frequently
Varnish
varnish-bugs at varnish-cache.org
Mon Dec 20 18:02:14 CET 2010
#837: varnishadm purge.list fails frequently
----------------------+-----------------------------------------------------
Reporter: jelder | Owner: kristian
Type: defect | Status: new
Priority: normal | Milestone: After Varnish 2.1
Component: varnishd | Version: trunk
Severity: normal | Keywords: purge.list
----------------------+-----------------------------------------------------
Comment(by jelder):
I've disabled the secret in one of the varnish instances in our staging
environment. I will reply to the ticket once we've reached the absurd
level of purges in production, and can verify that the issue is
reproducible via telnet.
We rely on saint mode (obj.grace = 1h), and as I understand things,
enabling the ban lurker would purge objects from the cache which we might
want to serve during an outage. Unless I misunderstand things, we would
need ban lurker to be aware of the grace period in some way, right?
{{{
sub vcl_recv {
set req.backend = haproxy;
if (req.backend.healthy) {
/* A different thread has already made a request to the backend
but it
* hasn't completed yet. */
set req.grace = 30s;
} else {
/* Either HAProxy is down or it's telling us via /haproxy_test
that all
* of its backends are down. */
set req.grace = 1h;
}
}
sub vcl_fetch {
/* Enter saint mode when we see at least backend.saintmode_threshold
of the
* following errors in a period of 10 seconds. */
if (
beresp.status == 500 || /* Internal Server Error */
beresp.status == 502 || /* Bad Gateway */
beresp.status == 503 || /* Service Unavailable */
beresp.status == 504 /* Gateway Timeout */
) {
set beresp.saintmode = 10s;
restart;
}
set beresp.grace = 1h;
}
}}}
--
Ticket URL: <http://www.varnish-cache.org/trac/ticket/837#comment:2>
Varnish <http://varnish-cache.org/>
The Varnish HTTP Accelerator
More information about the varnish-bugs
mailing list