Hi, Lately, on some BSDi-based (4.1 BSDI BSD/OS 4.1 i386) squids I help support, we've been seeing dramatic slowdown in squid performance. From one of the afflicted systems' cachemgr 60min report: client_http.hits = 18.376455/sec client_http.errors = 0.013243/sec client_http.kbytes_in = 81.341042/sec client_http.kbytes_out = 1117.131508/sec client_http.all_median_svc_time = 1.311657 seconds client_http.miss_median_svc_time = 1.542425 seconds client_http.nm_median_svc_time = 0.399283 seconds client_http.nh_median_svc_time = 1.242674 seconds client_http.hit_median_svc_time = 0.469653 seconds Normally, the median SVC time is much lower, like a full second lower. Coinciding with this, we've been seeing this in the cache.log: 2005/06/09 21:08:49| idnsSendQuery: FD 5: sendto: (55) No buffer space available 2005/06/09 21:08:49| comm_udp_sendto: FD 5, 192.168.1.2, port 53: (55) No buffer space available These errors aren't seen during non-business hours; only under "normal" load. I've Googled and searched the list, and haven't found anything exactly like this. Further, I'm relatively new to supporting squid, so I don't know exactly what this means. Any help is appreciated. Doug
