I've done some google searching on this already, but I don't see an answer, just questions about it. While trying to track down this issue where all cache store writes appear to work fine (the cache_dir is working as far as I can tell) but subsequent attempts to read the files out of cache fail, all of them do, greatly increasing the time pages take to load, and in some cases, the fail to load at all. When they do fail to finish loading (they usually start, but fail in the middle) hitting refresh in the browser sometimes works to finish loading the page as well. This problem started for me when I restarted the server for the first time in about a year, a few weeks ago. This is what is in my cache.log: (debug ALL,2) 2007/07/20 10:53:21| The reply for GET http://www.google.com/intl/en_uk/images/t4_ie.gif is ALLOWED, because it matched 'all' 2007/07/20 10:53:21| clientReadRequest: FD 67: no data to process ((11) Resource temporarily unavailable) (debug ALL,9) 2007/06/25 16:46:01| peerSourceHashSelectParent: Calculating hash for 10.0.0.20 2007/06/25 16:46:01| storeCreate: Selected dir '0' for obj size '2122' 2007/06/25 16:46:01| clientReadRequest: FD 20: no data to process ((11) Resource temporarily unavailable) 2007/06/25 16:46:01| The request GET http://img2.fark.com/images/2007/site/bodyGradientGeek.gif is ALLOWED, because it matched 'impromed' 2007/06/25 16:46:01| peerSourceHashSelectParent: Calculating hash for 10.0.0.20 2007/06/25 16:46:02| The reply for GET http://img2.fark.com/images/2007/site/bodyGradientGeek.gif is ALLOWED, because it matched 'all' 2007/06/25 16:46:02| storeCreate: Selected dir '0' for obj size '647' 2007/06/25 16:46:02| clientReadRequest: FD 20: no data to process ((11) Resource temporarily unavailable) 2007/06/25 16:46:02| The reply for GET http://secure-us.imrworldwide.com/cgi-bin/m?rnd=1182807956048&ci=us-maxim&cg=fark&cc=1&sr=1280x1024&cd=32&lg=en-US&je=y&ck=y&tz=-5&si=http%3A//www.fark.com/tech/&rp=http%3A//www.fark.com/sports/ is ALLOWED, because it matched 'all' I'm currently running squid 2.6 STABLE13 against a 2.6.17 kernel on Fedora Core 5. Squid was built with the following configure options: --enable-storeio=ufs,aufs,coss,diskd --enable-removal-policies=lru,heap --disable-wccp --disable-wccpv2 --enable-kill-parent-hack --with-dl --disable-internal-dns (the logs from 6/25 were from 2.6 STABLE12 that was built with just configure --prefix=/usr/local/squid) I've tried various ufs, aufs, diskd and single/multiple cache_dir and heap LRU / heap LFUDA / heap GDSF and plain lru configurations without fixing this. Thanks for any help you can give me on this matter, as conserving some bandwidth while not making using proxy slower than going direct is my goal. BTW: the connection is 10Mbit/s up/down over fiber if that makes a difference. Scott Anderson sbanderson@xxxxxxxxxxxx IT Administrator ImproMed, Inc.
