Hello, On apache 2.2.22, 2.2.23 and 2.4.4 I'm able to reproduce a problem where graceful restart takes very long time. Linux 3.7.10, glibc 2.17 here.
keepalive is off, ServerLimit 600; StartServers 64; MinSpareServers 32; MaxSpareServers 64; MaxClients 600; MaxRequestsPerChild 2000; minimal number of apache modules I'm running ab where 192.168.1.121 is my local ip: /usr/bin/ab -n 100000 -c 64 http://192.168.1.121/ (ususally I'm interrupting ab in first 1-2 seconds of its run, starting it again and leaving running) and then doing graceful restart while ab is still running: [Thu Apr 11 13:27:32.624875 2013] [mpm_prefork:notice] [pid 10202] AH00171: Graceful restart requested, doing restart [Thu Apr 11 13:28:15.836068 2013] [mpm_prefork:notice] [pid 10202] AH00163: Apache/2.4.4 (Unix) configured -- resuming normal operations It takes almost 1 minute to resume operations. Now if I don't do ab "magic" then resume is immediate (1-2s). The whole time is spend in prefork ap_mpm_pod_killpg ap_log_error(APLOG_MARK, APLOG_NOTICE, 0, ap_server_conf, APLOGNO(00171) "Graceful restart requested, doing restart"); /* kill off the idle ones */ ap_mpm_pod_killpg(pod, retained->max_daemons_limit); where ap_mpm_pod_killpg iterates over childrens (?) and that takes ~1 minute. I also did testing with StartServers 128. ap_mpm_pod_killpg had to iterate over 128 childrens where 72 of them were done in the same second while the rest in over minute - log here http://pastebin.com/3tP5hW1f. Example strace of main httpd process while doing graceful restart: http://pastebin.com/QFH5TjT6 The whole problem is that until ap_mpm_pod_killpg finishes apache is not responding to new requests and the downtime is between 30s to even 4 minutes. -- Arkadiusz Miśkiewicz, arekm / maven.pl
