<div dir="auto">Hi Igor,<div dir="auto"><br></div><div dir="auto">Thanks for the answer.</div><div dir="auto"><br></div><div dir="auto">I really got that message 'signal process started' every time i do 'nginx -s reload' and this is the only log line I have, I don't have the other lines you mentioned. Is there anything to do to enable those logs?</div><div dir="auto"><br></div>Sébastien<br><br><div class="gmail_quote" dir="auto"><div dir="ltr" class="gmail_attr">Le jeu. 28 mars 2024, 16:40, Igor Ippolitov <<a href="mailto:iippolitov@nginx.com">iippolitov@nginx.com</a>> a écrit :<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><u></u>

  
    
  
  <div>
    <div>Sébastien,</div>
    <div><br>
    </div>
    <div>The message about the signal process is
      only the beginning of the process.</div>
    <div>You are interested in messages like the
      following:<br>
      <br>
      <span style="font-family:monospace">
        <blockquote type="cite"><span style="font-family:monospace"><span style="color:#000000;background-color:#ffffff">2024/03/26
              13:36:36 [notice] 723#723: signal 1 (SIGHUP) received from
              69064, reconfiguring </span><br>
            2024/03/26 13:36:36 [notice] 723#723: reconfiguring <br>
            2024/03/26 13:36:36 [notice] 723#723: using the "epoll"
            event method <br>
            2024/03/26 13:36:36 [notice] 723#723: start worker processes
            <br>
            2024/03/26 13:36:36 [notice] 723#723: start worker process
            69065 <br>
            2024/03/26 13:36:36 [notice] 723#723: start worker process
            69066 <br>
            2024/03/26 13:36:36 [notice] 723#723: start cache manager
            process 69067 <br>
            2024/03/26 13:36:36 [notice] 61903#61903: gracefully
            shutting down <br>
            2024/03/26 13:36:36 [notice] 61905#61905: exiting <br>
            2024/03/26 13:36:36 [notice] 61903#61903: exiting <br>
            2024/03/26 13:36:36 [notice] 61904#61904: gracefully
            shutting down <br>
            2024/03/26 13:36:36 [notice] 61904#61904: exiting <br>
            2024/03/26 13:36:36 [notice] 61903#61903: exit</span></blockquote>
        <br>
      </span></div>
    <div>Note the 'gracefully shutting down' and
      'exiting' message from workers. Also the 'start' and
      'reconfiguring' messages from the master process.</div>
    <div>There should be a similar sequence
      somewhere in your logs.</div>
    <div>Having these logs may help explaining
      what happens on a reload.<br>
    </div>
    <div><br>
    </div>
    <div>Kind regards,</div>
    <div>Igor.<br>
    </div>
    <div><br>
    </div>
    <div>On 26/03/2024 12:41, Sébastien Rebecchi
      wrote:<br>
    </div>
    <blockquote type="cite">
      
      <div dir="ltr">Hi Igor
        <div><br>
        </div>
        <div>There is no special logs on the IP_1 (the reloaded one)
          side, only 1 log line, which is expected:</div>
        <div>--- BEGIN ---<br>
        </div>
        <div>2024/03/26 13:37:55 [notice] 3928855#0: signal process
          started<br>
        </div>
        <div>--- END ---<br>
        </div>
        <div><br>
        </div>
        <div>I did not configure worker_shutdown_timeout, it is
          unlimited.</div>
        <div><br>
        </div>
        <div>Sébastien.</div>
      </div>
      <br>
      <div class="gmail_quote">
        <div dir="ltr" class="gmail_attr">Le lun. 25 mars 2024 à 17:59,
          Igor Ippolitov <<a href="mailto:iippolitov@nginx.com" target="_blank" rel="noreferrer">iippolitov@nginx.com</a>>
          a écrit :<br>
        </div>
        <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
          <div>
            <div>Sébastien,</div>
            <div><br>
            </div>
            <div>Nginx should keep active connections open and wait for
              a request to complete before closing.<br>
            </div>
            <div>A reload starts a new set of workers while old workers
              wait for old connections to shut down.</div>
            <div>The only exception I'm aware of is having
              worker_shutdown_timeout configured: in this case a worker
              will wait till this timeout and forcibly close a
              connection. Be default there is no timeout.<br>
            </div>
            <div><br>
            </div>
            <div>It would be curious to see error log of nginx at IP_1
              (the reloaded one) while the reload happens. It may
              explain the reason for connection resets.<br>
            </div>
            <div><br>
            </div>
            <div>Kind regards,</div>
            <div>Igor.<br>
            </div>
            <div><br>
            </div>
            <div>On 25/03/2024 12:31, Sébastien Rebecchi wrote:<br>
            </div>
            <blockquote type="cite">
              <div dir="ltr">
                <div style="padding:0px 0px 0px 2px">
                  <div style="color:rgb(0,0,0);font-family:monospace;font-size:10pt">
                    <p style="margin:0px">Hello</p>
                    <p style="margin:0px"><br>
                    </p>
                    <p style="margin:0px">I have an issue with <span style="text-decoration-line:underline;text-decoration-style:wavy;text-decoration-color:rgb(255,128,64)">nginx</span>
                      closing prematurely connections when reload is
                      performed.</p>
                    <p style="margin:0px"><br>
                    </p>
                    <p style="margin:0px">I have some <span style="text-decoration-line:underline;text-decoration-style:wavy;text-decoration-color:rgb(255,128,64)">nginx</span>
                      servers configured to proxy_pass requests to an
                      upstream group. This group itself is composed of
                      several servers which are <span style="text-decoration-line:underline;text-decoration-style:wavy;text-decoration-color:rgb(255,128,64)">nginx</span>
                      themselves, and is configured to use <span style="text-decoration-line:underline;text-decoration-style:wavy;text-decoration-color:rgb(255,128,64)">keepalive</span>
                      connections.</p>
                    <p style="margin:0px">When I trigger a reload (-s
                      reload) on an <span style="text-decoration-line:underline;text-decoration-style:wavy;text-decoration-color:rgb(255,128,64)">nginx</span>
                      of one of the servers which is target of the
                      upstream, I see in error logs of all servers in
                      front that connection was reset by the <span style="text-decoration-line:underline;text-decoration-style:wavy;text-decoration-color:rgb(255,128,64)">nginx</span>
                      which was reloaded.</p>
                    <p style="margin:0px"><br>
                    </p>
                    <p style="margin:0px">Here configuration of upstream
                      group (IPs are hidden replaced by IP_X):</p>
                    <p style="margin:0px">--- BEGIN ---</p>
                    <p style="margin:0px">upstream data_api {</p>
                    <p style="margin:0px"> random;</p>
                    <p style="margin:0px"><br>
                    </p>
                    <p style="margin:0px"> server IP_1:80 max_fails=3
                      fail_timeout=30s;</p>
                    <p style="margin:0px"> server IP_2:80 max_fails=3
                      fail_timeout=30s;</p>
                    <p style="margin:0px"> server IP_3:80 max_fails=3
                      fail_timeout=30s;</p>
                    <p style="margin:0px"> server IP_4:80 max_fails=3
                      fail_timeout=30s;</p>
                    <p style="margin:0px"> server IP_5:80 max_fails=3
                      fail_timeout=30s;</p>
                    <p style="margin:0px"> server IP_6:80 max_fails=3
                      fail_timeout=30s;</p>
                    <p style="margin:0px"> server IP_7:80 max_fails=3
                      fail_timeout=30s;</p>
                    <p style="margin:0px"> server IP_8:80 max_fails=3
                      fail_timeout=30s;</p>
                    <p style="margin:0px"> server IP_9:80 max_fails=3
                      fail_timeout=30s;</p>
                    <p style="margin:0px"> server IP_10:80 max_fails=3
                      fail_timeout=30s;</p>
                    <p style="margin:0px"><br>
                    </p>
                    <p style="margin:0px"> <span style="text-decoration-line:underline;text-decoration-style:wavy;text-decoration-color:rgb(255,128,64)">keepalive</span>
                      20;</p>
                    <p style="margin:0px">}</p>
                    <p style="margin:0px">--- END ---</p>
                    <p style="margin:0px"><br>
                    </p>
                    <p style="margin:0px">Here configuration of the
                      location using this upstream:</p>
                    <p style="margin:0px">--- BEGIN ---</p>
                    <p style="margin:0px">location / {</p>
                    <p style="margin:0px"> proxy_pass <a href="http://data_api" target="_blank" rel="noreferrer">http://data_api</a>;</p>
                    <p style="margin:0px"><br>
                    </p>
                    <p style="margin:0px"> proxy_http_version 1.1;</p>
                    <p style="margin:0px"> proxy_set_header Connection
                      "";</p>
                    <p style="margin:0px"><br>
                    </p>
                    <p style="margin:0px"> proxy_set_header Host $host;</p>
                    <p style="margin:0px"> proxy_set_header X-Real-IP
                      $real_ip;</p>
                    <p style="margin:0px"> proxy_set_header
                      X-Forwarded-For $proxy_add_x_forwarded_for;</p>
                    <p style="margin:0px"><br>
                    </p>
                    <p style="margin:0px"> proxy_connect_timeout 2s;</p>
                    <p style="margin:0px"> proxy_send_timeout 6s;</p>
                    <p style="margin:0px"> proxy_read_timeout 10s;</p>
                    <p style="margin:0px"><br>
                    </p>
                    <p style="margin:0px"> proxy_next_upstream error
                      timeout <span style="text-decoration-line:underline;text-decoration-style:wavy;text-decoration-color:rgb(255,128,64)">http</span>_502
                      <span style="text-decoration-line:underline;text-decoration-style:wavy;text-decoration-color:rgb(255,128,64)">http</span>_504;</p>
                    <p style="margin:0px">}</p>
                    <p style="margin:0px">--- END ---</p>
                    <p style="margin:0px"><br>
                    </p>
                    <p style="margin:0px">And here the kind of error
                      messages I get when I reload <span style="text-decoration-line:underline;text-decoration-style:wavy;text-decoration-color:rgb(255,128,64)">nginx</span>
                      of "IP_1":</p>
                    <p style="margin:0px">--- BEGIN ---</p>
                    <p style="margin:0px">2024/03/25 11:24:25 [error]
                      3758170#0: *1795895162 <span style="text-decoration-line:underline;text-decoration-style:wavy;text-decoration-color:rgb(255,128,64)">recv</span>()
                      failed (104: Connection reset by peer) while
                      reading response header from upstream, client:
                      CLIENT_IP_HIDDEN, server: SERVER_HIDDEN, request:
                      "POST /REQUEST_LOCATION_HIDDEN HTTP/2.0",
                      upstream: "<a href="http://IP_1:80/REQUEST_LOCATION_HIDDEN" target="_blank" rel="noreferrer">http://IP_1:80/REQUEST_LOCATION_HIDDEN</a>",
                      host: "HOST_HIDDEN", referrer: "REFERRER_HIDDEN"</p>
                    <p style="margin:0px">--- END ---</p>
                    <p style="margin:0px"><br>
                    </p>
                    <p style="margin:0px">I thought -s reload was doing
                      graceful shutdown of connections. Is it due to the
                      fact that <span style="text-decoration-line:underline;text-decoration-style:wavy;text-decoration-color:rgb(255,128,64)">nginx</span>
                      can not handle that when using <span style="text-decoration-line:underline;text-decoration-style:wavy;text-decoration-color:rgb(255,128,64)">keepalive</span>
                      connections? Is it a bug?</p>
                    <p style="margin:0px">I am using <span style="text-decoration-line:underline;text-decoration-style:wavy;text-decoration-color:rgb(255,128,64)">nginx</span>
                      1.24.0 everywhere, no particular </p>
                    <p style="margin:0px"><br>
                    </p>
                    <p style="margin:0px">Thank you for any help.</p>
                    <p style="margin:0px"><br>
                    </p>
                    <p style="margin:0px"><span style="text-decoration-line:underline;text-decoration-style:wavy;text-decoration-color:rgb(255,128,64)">Sébastien</span></p>
                  </div>
                </div>
              </div>
              <br>
              <fieldset></fieldset>
              <pre>_______________________________________________
nginx mailing list
<a href="mailto:nginx@nginx.org" target="_blank" rel="noreferrer">nginx@nginx.org</a>
<a href="https://mailman.nginx.org/mailman/listinfo/nginx" target="_blank" rel="noreferrer">https://mailman.nginx.org/mailman/listinfo/nginx</a>
</pre>
            </blockquote>
            <p><br>
            </p>
          </div>
        </blockquote>
      </div>
    </blockquote>
    <p><br>
    </p>
  </div>

</blockquote></div></div>