Hello!

On Thu, Jul 05, 2018 at 07:29:37PM +0500, Илья Шипицин wrote:

[...]

> # cat /var/log/nginx/error.log | tail -2
> 2018/07/05 14:27:53 [alert] 1546#1546: 512 worker_connections are not enough
> 2018/07/05 14:27:53 [alert] 1545#1545: worker process 1546 exited with
> fatal code 2 and cannot be respawned
> #
>
>
> кажется, что было бы логично отсекать такие ошибки во время "nginx -t"
> что думаете ?

Кажется, ситуация не отличается принципиально от любой другой,
когда nginx'у не хватает соединений для нормальной работы. Но
можно попробовать какой-то такой патч, сколько-то ног он наверное
сохранит:

# HG changeset patch
# User Maxim Dounin <mdounin@mdounin.ru>
# Date 1530840179 -10800
# Fri Jul 06 04:22:59 2018 +0300
# Node ID 7f4aa03a8a21164881429568bd2f70a311c5c599
# Parent 54683f650cbdcd73f7f8d845c843295978da5a85
Events: added configuration check on the number of connections.

There should be at least one worker connection for each listening socket,
plus an additional connection for channel between worker and master,
or starting worker process will fail.

diff --git a/src/event/ngx_event.c b/src/event/ngx_event.c
--- a/src/event/ngx_event.c
+++ b/src/event/ngx_event.c
@@ -416,6 +416,21 @@ ngx_event_init_conf(ngx_cycle_t *cycle,
return NGX_CONF_ERROR;
}

+ if (cycle->connection_n < cycle->listening.nelts + 1) {
+
+ /*
+ * there should be at least one connection for each listening
+ * socket, plus one connection for channel
+ */
+
+ ngx_log_error(NGX_LOG_EMERG, cycle->log, 0,
+ "%ui worker_connections are not enough "
+ "for nginx with %ui listening sockets",
+ cycle->connection_n, cycle->listening.nelts);
+
+ return NGX_CONF_ERROR;
+ }
+
return NGX_CONF_OK;
}



--
Maxim Dounin
http://mdounin.ru/
_______________________________________________
nginx-ru mailing list
nginx-ru@nginx.org
http://mailman.nginx.org/mailman/listinfo/nginx-ru