2010/2/17 Henrik Nordström <henrik@xxxxxxxxxxxxxxxxxxx>: > tor 2010-02-18 klockan 14:51 +1300 skrev Amos Jeffries: > >> Henrik seems to have re-appeared and he has more disk IO experience then >> me so may have an idea whet to look for ... ? > > My first reaction is to run a small benchmark in parallel to squid > performing a reasonably slow sequence of random reads over the disks and > measuring the response time.. SSD disks in particular may have a habit > of "randomly" block all requests for a while during/after a period of > writes.. and OS write buffering and batching may even add to this queue > latency problem. > > But I have not read up on the whole problem description. > > Also keep in mind that as Amos mentioned earlier these "Queue > congestion" warnings come more frequently after start as there is a > filter preventing the logs from getting overwhelmed with this warning. Thanks Henrik and Amos I'll do whatever testing is needed. As it's really odd.and I can't use these servers/disks until I can get this problem identified. And sorry "sleeping" was just my way of citing the box shows no load, almost no IO 4-5 when I'm hitting it hard. I do not see this issue with lesser threads, it's only when I turn up the juice. But with turning up the connections per second I would expect to see some type of load and I see none. and the -X -d I dont see anything but that error, is there another log file that needs to be enabled vs cache.log and squid.out? I'll try some various things to see what I can see. I know time dd tests and bonnie++ and some FIO seem to do just fine. It's only squid that seems to be having an issue with me or my setup :) Tory
