При переключении fastcgi с tcp коннекта (fastcgi_pass 127.0.0.1:9000;) на сокеты (fastcgi_pass unix:/tmp/php-fpm.sock;) и появилась проблема... При загрузке файлов иногда (!) стала выдаваться 413 ошибка... в логах получаем следующее:
2010/06/27 10:55:50 [error] 44923#0: *309634 upstream timed out (60: Operation timed out) while sending request to upstream, client: 111.77.222.111, server: domain.ru, request: "POST /main/save/ HTTP/1.1", upstream: "fastcgi://unix:/tmp/php-fpm.sock:", host: "domain.ru"
2010/06/27 11:06:49 [error] 44922#0: *310300 upstream timed out (60: Operation timed out) while sending request to upstream, client: 111.77.222.111, server: domain.ru, request: "POST /main/saveTracks/?id=1261 HTTP/1.1", upstream: "fastcgi://unix:/tmp/php-fpm.sock:", host: "domain.ru"
2010/06/27 11:06:49 [error] 44922#0: *310300 client intended to send too large body: 8364210 bytes, client: 111.77.222.111, server: domain.ru, request: "POST /main/saveTracks/?id=1261 HTTP/1.1", upstream: "fastcgi://unix:/tmp/php-fpm.sock:", host: "domain.ru"
2010/06/27 11:07:59 [error] 44923#0: *310370 upstream timed out (60: Operation timed out) while sending request to upstream, client: 111.77.222.111, server: domain.ru, request: "POST /main/saveTracks/?id=1261 HTTP/1.1", upstream: "fastcgi://unix:/tmp/php-fpm.sock:", host: "domain.ru"
2010/06/27 11:07:59 [error] 44923#0: *310370 client intended to send too large body: 1318466 bytes, client: 111.77.222.111, server: domain.ru, request: "POST /main/saveTracks/?id=1261 HTTP/1.1", upstream: "fastcgi://unix:/tmp/php-fpm.sock:", host: "domain.ru"
2010/06/27 11:09:04 [error] 44923#0: *310467 upstream timed out (60: Operation timed out) while sending request to upstream, client: 111.77.222.111, server: domain.ru, request: "POST /main/saveTracks/?id=1261 HTTP/1.1", upstream: "fastcgi://unix:/tmp/php-fpm.sock:", host: "domain.ru"
2010/06/27 11:09:04 [error] 44923#0: *310467 client intended to send too large body: 8791573 bytes, client: 111.77.222.111, server: domain.ru, request: "POST /main/saveTracks/?id=1261 HTTP/1.1", upstream: "fastcgi://unix:/tmp/php-fpm.sock:", host: "domain.ru"
2010/06/27 11:10:31 [error] 44923#0: *310572 upstream timed out (60: Operation timed out) while sending request to upstream, client: 111.77.222.111, server: domain.ru, request: "POST /main/saveTracks/?id=1261 HTTP/1.1", upstream: "fastcgi://unix:/tmp/php-fpm.sock:", host: "domain.ru"
2010/06/27 11:10:31 [error] 44923#0: *310572 client intended to send too large body: 14275204 bytes, client: 111.77.222.111, server: domain.ru, request: "POST /main/saveTracks/?id=1261 HTTP/1.1", upstream: "fastcgi://unix:/tmp/php-fpm.sock:", host: "domain.ru"
Срабатывает, по всей видимости, emergency_restart_interval php-fpm'а.(?) Переключая обратно, получаем отлично работающий механизм. Что делать и куда копать уже не знаю, перерыл за две недели много, всё обновлял до последних версий и переставлял заново - эффект нулевой...