Awesome, this worked great, thanks!
The one caveat that others might run into was that sometimes the backend
would sent / as the url, and I was using the regexp ~ version from the
link you sent, so it would clear the whole cache. I changed it to this:
purge("req.url ~ " req.url "$"); # use regexp to
handle all Accept-Encoding's
error 200 "Purged in recv.";
and now it works perfectly.
Chris
On 2010/07/09 04:56, Laurence Rowe wrote:
On 9 July 2010 12:01, Chris Hecker<[email protected]> wrote:
I'm trying to put varnish in front of mediawiki, but there are a couple
things going wrong with the purging support built into mediawiki (which
supposedly supports varnish and squid).
1. Mediawiki was sending back an HTTP/1.0 PURGE command, which had no Host:
header, just the full url after the PURGE, so the hash function wasn't
matching with the normal HTTP/1.1 requests. I patched mediawiki to fix this
and send an HTTP/1.1 PURGE with a Host: header, not sure what a better way
to do this would be.
2. The PURGE does not have an Accept-encoding: on it, so it doesn't match
the cached object, which is almost always A-e: gzip. I hacked a gzip header
onto the PURGE, but what I really want is to purge all the records
associated with an URL. Is there a way to do this?
See: http://varnish-cache.org/wiki/VCLExamplePurging
Using purge("req.url == " req.url); should fix both problems.
Laurence
_______________________________________________
varnish-misc mailing list
[email protected]
http://lists.varnish-cache.org/mailman/listinfo/varnish-misc
_______________________________________________
varnish-misc mailing list
[email protected]
http://lists.varnish-cache.org/mailman/listinfo/varnish-misc
