with squid 2.6 we got many TCP Connection dmz-proxy failed messages. these messages disappear with squid 2.5. last week henrik nordstrom answered my questions about those messages, that squid 2.6 performs much better under load. so this will explain why with 2.6 the messages show up in cache.log and with 2.5 not. i wonder what we can optimize further to stop those errors. there is quite a lot traffic, our dmz proxy handles at about 6 Mio hits per day, our "internal"-proxy handles at about 10 Mio hits per day. not really sure whether one or both squids are on their limits? uptime says dmz: load average: 0.70, 0.86, 0.84 internal: load average: 1.58, 2.01, 1.95 both are "quad"-processor systems, with 2 GB ram... so i wonder what we can optimize so that our squids perform better? one planned option is to do load balancing with two proxy-servers (both internal/dmz), but this will be at the end of year, matter of budget and planned hardware ordering... >-----Ursprüngliche Nachricht----- >Von: Adrian Chadd [mailto:adrian@xxxxxxxxxxxxxxx] >Gesendet: Montag, 21. August 2006 10:35 >An: Rietzler, Markus (Firma Rietzler Software / RZF) >Cc: squid-users@xxxxxxxxxxxxxxx >Betreff: Re: cache.log: got unsused STORE_META type 10 > >On Mon, Aug 21, 2006, Markus.Rietzler@xxxxxxxxxxxxxx wrote: >> but: there are quite a lot, this morning 220.000(!) lines. >> in the same time we had at about 5.000.000 hits on our cache, >> so 4% of requests lead to the message... > >Upgrading to the latest squid-2.5 stable release will silence these >messages. :) > >What was weird with squid-2.6 which made you downgrade? > > >Adrian >
