On 30 April 2012 12:49, Neil Saunders <[email protected]> wrote: > Hi all - > > I'm attempting to implement a suggestion provided as a solution to another > question I posted regarding cache warming. > > Long story short, I have 6 webservers that I'm pre-warming 60,000 urls via a > script. I had previously been sending each request to each web-server, but > it was suggested that it would be much quicker, and indeed more elegant, to > be able to set a header (X-Cache-Warming in this case) that if set would > cause the web-server to use the next web server as the backend, until it > reached the last web server and it would be fetched via the actual backend: > Goal - Make a single request on the first web server to warm all 6. > > The issue I'm seeing is following cache warming the I get cache misses on > actual requests on all web servers except the last in the chain, which would > imply to me that Varnish implicitly hashes the backend name used.
I don't have an answer for this, but here's a thought: try explicitly logging the "req.hash" value as late as possible in the vcl_* chain (I don't know when/where it's an acceptable variable to query) to see what it produces. https://www.varnish-cache.org/docs/3.0/reference/vcl.html#the-hash-director says it uses just this as the key. I don't know if it contains an opaque lookup key, or something more useful. Perhaps comparing a single request's req.hash value between multiple chained caches in your setup will show something interesting ... Jonathan -- Jonathan Matthews Oxford, London, UK http://www.jpluscplusm.com/contact.html _______________________________________________ varnish-misc mailing list [email protected] https://www.varnish-cache.org/lists/mailman/listinfo/varnish-misc
