sendfile_max_chunk breaking unbufferred php-fcgi

[Valentin V. Bartenev]

On Sunday 14 June 2015 09:52:26 Floris wrote:
> Hi,
> 
> I was having the problem that if a single client on the local LAN is
> downloading a large static file, the download is effectively monopolizing
> nginx, and no other requests are handled simultaneously.
> Reading the manual I came across the sendfile_max_chunk option that sounded
> like that may fix it:
> 
> ==
> Syntax:    sendfile_max_chunk size;
> Default:    sendfile_max_chunk 0;
> Context:    http, server, location
> 
> When set to a non-zero value, limits the amount of data that can be
> transferred in a single sendfile() call. Without the limit, one fast
> connection may seize the worker process entirely.
> ==
> 
> 
> However I noticed that if I enable that, PHP scripts running without
> buffering suddenly no longer work properly.
> 
> 
> nginx.conf:
> 
> ==
> events {
>     worker_connections  1024;
> }
> 
> http {
>     include      mime.types;
>     default_type application/octet-stream;
> 
>     server {
>         listen      80;
>         server_name $hostname;
>         sendfile    on;
>         sendfile_max_chunk 8192;
> 
>         root /var/www;
> 
>         location / {
>             index index.php index.html index.htm;
>         }
> 
>         location ~ \.php$ {
>             try_files     $uri =404;
> 
>             fastcgi_buffering off;
>             fastcgi_pass  unix:/var/run/php-fpm.sock;
>             include       fastcgi.conf;
>         }
>     }
> }
> ==
> 
> 
> t2.php for testing purposes:
> 
> ==
> <?php
> 
> for ($i = 0; $i < 10; $i++)
> {
>         echo "test!\n";
>         flush();
>         sleep(1);
> }
> ==
> 
> When retrieving that, the connection stalls after the first flush:
> 
> ==
> $ telnet 192.168.178.26 80
> Trying 192.168.178.26...
> Connected to 192.168.178.26.
> Escape character is '^]'.
> GET /t2.php HTTP/1.0
> 
> HTTP/1.1 200 OK
> Server: nginx/1.6.3
> Date: Sun, 14 Jun 2015 13:21:53 GMT
> Content-Type: text/html; charset=UTF-8
> Connection: close
> X-Powered-By: PHP/5.6.9
> 
> test!
> ==
> 
> If I remove either the "sendfile_max_chunk 8192;" or "fastcgi_buffering
> off;" line it does work, and I do get all 10 test! messages:
> 
> ==
> telnet 192.168.178.26 80
> Trying 192.168.178.26...
> Connected to 192.168.178.26.
> Escape character is '^]'.
> GET /t2.php HTTP/1.0
> 
> HTTP/1.1 200 OK
> Server: nginx/1.6.3
> Date: Sun, 14 Jun 2015 13:22:23 GMT
> Content-Type: text/html; charset=UTF-8
> Connection: close
> X-Powered-By: PHP/5.6.9
> 
> test!
> test!
> test!
> test!
> test!
> test!
> test!
> test!
> test!
> test!
> Connection closed by foreign host.
> ==
> 
> Am I doing something wrong, or is this a bug?
> 

Yes, it's a bug.

You should set "sendfile_max_chunk 0;" in the location with unbuffered fastcgi.

Or you can try the following patch:

diff -r c041f1e0655f src/http/ngx_http_upstream.c
--- a/src/http/ngx_http_upstream.c      Wed Jun 10 19:18:20 2015 +0300
+++ b/src/http/ngx_http_upstream.c      Mon Jun 15 13:32:55 2015 +0300
@@ -3303,6 +3303,9 @@ ngx_http_upstream_process_non_buffered_r
     downstream = r->connection;
     upstream = u->peer.connection;
 
+    /* workaround for sendfile_max_chunk */
+    downstream->write->delayed = 0;
+
     b = &u->buffer;
 
     do_write = do_write || u->length == 0;