My configuration is a bit specific. I have an Apache server for PHP and Python stuff with nginx as a frontend (and for serving static files like this blog). This part seems normal for most of people. BUT, in my case, I want that the two servers write the access logs IN THE SAME FILE (/var/log/apache2/access.log)

I run a Debian stable on my server. With the previous version (7.0), all was fine, but, until Jessie (8.0), I found a very strange behaviour with my web statistics analyzer. I usually have from 60 to 80 visits per day. In may (after the upgrade to Jessie), I get only 2 or 3 visits per day, using only one subdomain. Plus, it's a period where I have no internet at home and no time to investigate...

Finally, I found the problem ! It's located in logrotate. The two files are :

/etc/logrotate.d/apache2

/var/log/apache2/*.log {
    daily
    missingok
    rotate 14
...
    postrotate
            if /etc/init.d/apache2 status > /dev/null ; then \
                /etc/init.d/apache2 reload > /dev/null; \
            fi;
    endscript
...
}

And /etc/logrotate.d/nginx

/var/log/nginx/*.log {
    weekly
    missingok
    rotate 52
...
    postrotate
            invoke-rc.d nginx rotate >/dev/null 2>&1
    endscript
}

The first one do a logrotate every day of all files in /var/log/apache2/*.log (that contains our . The second one, do only one logrotate every week. But, the most important line is located in postrotate. Why it's important ? Because it says to nginx and Apache "Now the log file has changed, close the previous and open the new one". Seems correct, but without that, nginx write in /var/log/apache2/access.log.1 and not /var/log/apache2/access.log until it gets rotated (after a week) because it will keep the handle on access.log that is renamed to access.log.1 by logrotate.

The solution is to add

invoke-rc.d nginx rotate >/dev/null 2>&1

In the postrotate section of /etc/logrotate.d/apache2 (after the if). Then, restart nginx :

systemctl restart nginx