Hi BR,

I don't known how to reproduce, not yet :-)

I couldn't identify which worker was responding too, but I can see with strace warnings in the old wolker about EAGAIN (Resource temporarily unavailable). I can see that because old workers still running:

USER PID %CPU %MEM VSZ RSS TTY STAT START TIME COMMAND
nginx 6009 0.0 0.2 71020 18532 ? S Sep18 0:37 nginx: worker process is shutting down
nginx 6010 0.0 0.2 71028 18496 ? S Sep18 0:12 nginx: worker process is shutting down
nginx 6289 0.2 0.3 75672 23188 ? S 07:54 0:58 nginx: worker process is shutting down
nginx 6290 0.0 0.2 71932 19248 ? S 07:54 0:15 nginx: worker process is shutting down
nginx 9182 0.0 0.2 70872 18380 ? S 10:20 0:02 nginx: worker process is shutting down
nginx 9295 0.0 0.2 70952 18380 ? S 10:26 0:02 nginx: worker process is shutting down
nginx 9297 0.0 0.2 70368 17856 ? S 10:26 0:02 nginx: worker process is shutting down
nginx 9302 0.0 0.2 70804 18296 ? S 10:26 0:01 nginx: worker process is shutting down
nginx 10132 0.2 0.2 74776 22280 ? S 10:53 0:47 nginx: worker process is shutting down
nginx 10133 0.0 0.2 71484 18972 ? S 10:53 0:09 nginx: worker process is shutting down
nginx 13690 0.2 0.2 72876 20296 ? S 14:22 0:10 nginx: worker process
nginx 13691 0.1 0.2 71492 19088 ? S 14:22 0:07 nginx: worker process
nginx 13692 0.0 0.0 57292 3180 ? S 14:22 0:00 nginx: cache manager process
root 29863 0.0 0.0 57292 4048 ? Ss Sep11 0:00 nginx: master process /usr/sbin/nginx -c /etc/nginx/nginx.conf
nginx 30956 0.0 0.2 72924 20336 ? S Sep18 0:23 nginx: worker process is shutting down
[/code]

Looking for our user's usage, this workers will stay online for more few hours :)

The difference between old and new configuration is just a "down" flag in one of our servers from our websockets pool (upstream). You can see a simplified version of config on: http://pastebin.com/02GQQ22r

Really thanks for your attention.

Igor