On Thu, Jan 12, 2012 at 12:53:18PM +1300, Amos Jeffries wrote: Hi, > So doing I/O to a disk log somehow speeds up TCP throughput? strange definitely right ... but I'm not sure if this speeds up TCP throughput - but speeds up Squid ;-) > This sounds a bit like the speed problems we see with very low traffic > rates. When the I/O loops get very few requests through they end up > pausing in 10ms time chunks each processing cycle to prevent CPU overload > doing lots of processing on very small amounts of bytes. hmm, speed problems should be no problem. I tested also the following: - iptables -t nat -A PREROUTING -p tcp --dport 8081 -j DNAT --to-destination 212.112.181.17:80 - iptables -t nat -A POSTROUTING -p tcp -d 212.112.181.17 -j MASQUERADE no changes in client so every browser url gets redirected to the ment. website. Dirty but what I found out was same speed as a direct internet connection. So speed problems not given - btw: our internet access has speed of 1G :-) > This release is getting a bit old now and has a few I/O buffering bugs in > it that may be related. > Please try the 3.1.18 Debian package from Wheezy / testing repositories > (may require some dependency updates as well). installed testing release - no success :-( > 145 connect() calls in 0.05 ms, all failing? does not seem right at > all. you're right, all calls where connect for IPv6-addresses. But we have this failed calls also with other sites. Testing deactivated IPv6 in OS was decided based on these connect-calls. Anyway the connect calls still used IPv6-addresses. But we have these connect fails with other websites too. So I don't think it's not the root cause. > Given the time measure I don't think its related, but probably worth > knowing and fixing. Did the section 5 trace show what was going on here? hmm I didn't find anything helpful but there are lots of messages. I can provide complete log if needed ... > Add here: > refresh_pattern -i (/cgi-bin/|\?) 0 0% 0 configured > Eeek! nearly unlimited access to the whole Internet. Why? only for test purposes :-) Therefore I used tcp/8081, our customer uses 8080 and this config has some ACLs more ... I stripped down configuration to exclude configuration problems ... Another really interesting result from another test. In my home environment I have also a Squid vom Debian Squeeze. Running in a VZ but on 32 Bit environment with nearly same configuration - no problem! And only with a 12MBit internet access ... So should this be a 64bit-related problem? I can't believe ... regards, Andreas Schulz
