Hi, I'm running: Squid Cache: Version 3.5.15 (including patches up to revision 14000) on FreeBSD 9.3-STABLE (recently updated) Every week or so I run into a problem where squid's CPU usage starts growing slowly, reaching 100% over the course of a day or so. When running normally its CPU usage is usually less than 5%. Restarting squid fixes the problem. Memory usage is about 2 GBytes (on a system with 8 GBytes of RAM). The number of socket connections (from clients and to servers) is about the same (roughly 500) when I have the problem as when I don't have the problem. Attaching GDB and getting a stack trace while squid is stuck at 100% generally gives me this: #0 0x00000000005deef4 in mem_node::end () #1 0x00000000005df076 in mem_node::dataRange () #2 0x0000000000625d34 in mem_hdr::NodeCompare () #3 0x0000000000628ad1 in SplayNode<mem_node*>::splay<mem_node*> () #4 0x0000000000628b85 in Splay<mem_node*>::find<mem_node*> () #5 0x0000000000625f8e in mem_hdr::getBlockContainingLocation () #6 0x0000000000625ff8 in mem_hdr::hasContigousContentRange () #7 0x00000000005e00fe in MemObject::isContiguous () #8 0x0000000000649d05 in StoreEntry::mayStartSwapOut () #9 0x0000000000648b96 in StoreEntry::swapOut () #10 0x0000000000639e87 in StoreEntry::invokeHandlers () #11 0x0000000000633e09 in StoreEntry::write () #12 0x000000000079caa1 in Client::storeReplyBody () #13 0x000000000059c0bf in HttpStateData::writeReplyBody () #14 0x00000000005a18fd in HttpStateData::processReplyBody () #15 0x00000000005a41ce in HttpStateData::processReply () #16 0x00000000005a4408 in HttpStateData::readReply () #17 0x00000000005ab6df in JobDialer<HttpStateData>::dial () #18 0x00000000006fd81a in AsyncCall::make () #19 0x0000000000701bc6 in AsyncCallQueue::fireNext () #20 0x0000000000701ecf in AsyncCallQueue::fire () #21 0x0000000000566621 in EventLoop::dispatchCalls () #22 0x0000000000566930 in EventLoop::runOnce () #23 0x0000000000566b18 in EventLoop::run () #24 0x00000000005dbb73 in SquidMain () #25 0x00000000005dc0fd in SquidMainSafe () #26 0x00000000004cf401 in _start () #27 0x0000000800ae4000 in ?? () #28 0x0000000000000000 in ?? () The cache.log file gets a few lines looking like this: 2016/03/31 11:51:04 kid1| local=192.168.1.15:3128 remote=192.168.1.164:49540 FD 339 flags=1: read/write failure: (60) Operation timed out and some others looking like this: 2016/03/31 14:40:05 kid1| FD 16, 192.168.1.15 [Stopped, reason:Listener socket closed job3132772]: (53) Software caused connection abort Does anybody have any suggestions on how to fix/improve this? Currently I have cron restarting squid every morning. Should I file a bug? Thanks Mark _______________________________________________ squid-users mailing list squid-users@xxxxxxxxxxxxxxxxxxxxx http://lists.squid-cache.org/listinfo/squid-users
