Plone: Varnish Configuration (Cache Hits, Purge Fails)

// under cache plone problem purge varnish

What a painful, painful day. It’s taken about the entire day, but I’ve finally figured out why our Varnish configuration around here for Plone/Zope hasn’t been accepting PURGE requests. Thus, without accepting purge requests correctly, any content that gets changed in Plone doesn’t get updated in the cache until it times out (or we restart it). Ouch - especially because it affects all sites under virtual hosting, and a lot of these sites have publicly visible content that needs to be available immediately.

So, the problem manifested itself as thus:

Cache hits happen
Performance was good on the Plone sites
PURGE requests sent to Varnish would produce 404’s

So, after checking (many times) the object path for the PURGE was right, Varnish would return a MISS and produce a 404 (DEBUG “VCL_error(404, Not in cache)” as the Varnish message).

Much searching/etc later, I found that using:

curl -X PURGE http://localhost:8000/object/path

straight on the machine running Varnish would work and give a 200 Purged response. But, running the same command (replacing localhost with a FQDN) from the Zope machine produced the 404. So, essentially, Varnish knows where to look for the first command but not for the second.

Solution

Where you set the backend for your Plone/Zope in the Varnish configuration, add set req.http.host accordingly:

} elseif (req.url ~ "^/VirtualHostBase/https?/my.plone.site.com:(80\|443)/plone/VirtualHostRoot")
{
set req.backend = zope1;
set req.http.host = "my.plone.site.com";
}
...

In short, the URL is processed from Apache (in my situation) to result in the req.url, and then Varnish needs to know what host anything being cached relates to. Without it, hello 404 heaven.

Go Top