Ubuntu php5-fpm throws unknown instance on reload

问题:

I am having some problems with ubuntu and php5-fpm on my VPS. Php works fine, however it throws terminal error when I am trying to reload / restart / stop it. Basically, I cannot stop php because it does not recognize the instance

It says "reload: Unknown instance:" If I try "service php5-fpm restart" it sais unknown instance and gives this in logs

[15-Nov-2013 11:56:40] ERROR: An another FPM instance seems to already listen on /var/run/php5-fpm.sock
[15-Nov-2013 11:56:40] ERROR: FPM initialization failed
[15-Nov-2013 11:56:40] NOTICE: configuration file /etc/php5/fpm/php-fpm.conf test is successful

I have php listening on /var/run/php5-fpm.sock ; I use nginx for webserver and it is correctly configured with

fastcgi_pass unix:/var/run/php5-fpm.sock;

(or so I know) Also need to mention that this happened after a reboot.

Site is working well, however I don't think this is a good sign. Any thoughts ? Thank you.

EDIT : I did as suggested in below post. It seems that if i kill the process, and then use start/stop it works fine. However, after I use reload command, it stops recongizing the instance.

root@developer2:/# service php5-fpm stop
stop: Unknown instance:
root@developer2:/# ps -ef
UID        PID  PPID  C STIME TTY          TIME CMD
root         1     0  0 Nov14 ?        00:00:00 init
root      1101     1  0 Nov14 ?        00:00:00 /sbin/udevd --daemon
root      1168     1  0 Nov14 ?        00:00:00 /usr/sbin/sshd -D
root      1227     1  0 Nov14 ?        00:00:00 /usr/sbin/xinetd -dontfork -pidfile /var/run/xinetd.pid -stayalive -inetd_compa
root      1229     1  0 Nov14 ?        00:00:00 cron
mysql     1249     1  0 Nov14 ?        00:00:18 /usr/sbin/mysqld
syslog    1283     1  0 Nov14 ?        00:00:00 /sbin/syslogd -u syslog
memcache  1293     1  0 Nov14 ?        00:00:07 /usr/bin/memcached -m 64 -p 11211 -u memcache -l 127.0.0.1
root      1426     1  0 Nov14 ?        00:00:00 /usr/lib/postfix/master
postfix   1440  1426  0 Nov14 ?        00:00:00 qmgr -l -t fifo -u
root      1468     1  0 Nov14 ?        00:00:00 /usr/sbin/varnishd -P /var/run/varnishd.pid -a :80 -T localhost:6082 -f /etc/va
nobody    1469  1468  0 Nov14 ?        00:01:14 /usr/sbin/varnishd -P /var/run/varnishd.pid -a :80 -T localhost:6082 -f /etc/va
root      1593     1  0 Nov14 ?        00:00:14 /usr/bin/python /usr/bin/fail2ban-server -b -s /var/run/fail2ban/fail2ban.sock
root      1595     1  0 Nov14 ?        00:00:00 /usr/lib/gamin/gam_server
www-data  3535     1  0 Nov14 ?        00:00:30 php-fpm: pool www
postfix  10016  1426  0 11:18 ?        00:00:00 pickup -l -t fifo -u -c
root     10064  1168  0 11:37 ?        00:00:00 sshd: root@pts/0
root     10080 10064  0 11:37 pts/0    00:00:00 -bash
root     10143  1168  0 11:47 ?        00:00:00 sshd: root@notty
root     10159 10143  0 11:47 ?        00:00:00 /usr/lib/openssh/sftp-server
root     11330     1  0 12:03 ?        00:00:00 nginx: master process /usr/sbin/nginx
www-data 11331 11330  0 12:03 ?        00:00:00 nginx: worker process
www-data 11332 11330  0 12:03 ?        00:00:00 nginx: worker process
www-data 11333 11330  0 12:03 ?        00:00:00 nginx: worker process
www-data 11334 11330  0 12:03 ?        00:00:00 nginx: worker process
root     11465  1168  0 12:14 ?        00:00:00 sshd: root@notty
root     11481 11465  0 12:14 ?        00:00:00 /usr/lib/openssh/sftp-server
root     11519 10080  0 12:23 pts/0    00:00:00 ps -ef
root@developer2:/# kill 3535
root@developer2:/# service php5-fpm start
php5-fpm start/running, process 11529
root@developer2:/# service php5-fpm stop
php5-fpm stop/waiting
root@developer2:/# service php5-fpm start
php5-fpm start/running, process 11544
root@developer2:/# service php5-fpm stop
php5-fpm stop/waiting
root@developer2:/# service php5-fpm start
php5-fpm start/running, process 11559
root@developer2:/# service php5-fpm reload
root@developer2:/# service php5-fpm reload
reload: Unknown instance:

Edit 2 : my php5 seems to create 3 processes

root     11813     1  1 12:34 ?        00:00:00 php-fpm: master process (/etc/php5/fpm/php-fpm.conf)
www-data 11815 11813  0 12:34 ?        00:00:00 php-fpm: pool www
www-data 11816 11813  0 12:34 ?        00:00:00 php-fpm: pool www

When I use reload function the master process dissapears and I have to manually kill the other 2 processes in order to be able to correctly start it again.

回答1:

This is a Ubuntu bug. In /etc/init/php5-fpm.conf, there is a commented-out line reload signal USR2, which causes the reload function to terminate the php5-fpm master process by sending SIGHUP. Any further reloads, restarts or stops will fail, because the master process has been terminated.

I fixed this in Ubuntu 14.04 by creating a file /etc/init/php5-fpm.override with the single line reload signal USR2 in it. Credits to Jurian Sluiman's comment in this answer.

Here is the primary bug report, proposed workarounds, and confirmation of workaround.

$ ps aux | grep php
USER       PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
root      1966  0.0  0.9 473276 37040 ?        Ss   10:03   0:00 php-fpm: master process (/etc/php5/fpm/php-fpm.conf)                    
www-data  2009  0.0  1.5 478280 62500 ?        S    10:03   0:01 php-fpm: pool www                                                       
www-data  2011  0.0  1.3 476504 55220 ?        S    10:03   0:00 php-fpm: pool www                                                       
www-data  2012  0.0  1.6 481592 65840 ?        S    10:03   0:00 php-fpm: pool www

$ sudo service php5-fpm status
php5-fpm start/running, process 1966

$ sudo service php5-fpm stop
php5-fpm stop/waiting

$ sudo service php5-fpm status
php5-fpm stop/waiting

$ sudo service php5-fpm start
php5-fpm start/running, process 2651

$ sudo service php5-fpm status
php5-fpm start/running, process 2651

$ ps aux | grep php
USER       PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
root      2651  0.2  0.9 473276 36996 ?        Ss   10:35   0:00 php-fpm: master process (/etc/php5/fpm/php-fpm.conf)                    
www-data  2654  0.0  0.1 473276  7104 ?        S    10:35   0:00 php-fpm: pool www                                                       
www-data  2655  0.0  0.1 473276  7104 ?        S    10:35   0:00 php-fpm: pool www                                                       
www-data  2656  0.0  0.1 473276  7104 ?        S    10:35   0:00 php-fpm: pool www

回答2:

Usually, the service command will do fine:

service php5-fpm restart

But if the "unknown instance" issue appears, you can just kill the processes and have the service restarted, using this line:

sudo pkill php5-fpm; sudo service php5-fpm start

回答3:

Try to stop your php5-fpm instance by service php5-fpm stop again. Wait for some seconds. Try to see, which processes was not terminated using ps -ef. Terminate them by kill <processId>. Start php5-fpm again. Is seems, not all child processes was terminated properly by service ... command. I have similar issues in on server, but they are not regular

回答4:

I have experienced the same issue using Envoyer for Laravel - and that was quite annoying. My solution was to follow option 3) at https://bugs.launchpad.net/ubuntu/+source/php5/+bug/1242376/comments/16

Create /etc/init/php5-fpm.override file with the single line "reload signal USR2" in it.

回答5:

Deploying with Envoyer brought me here. I ended up adding a 'before' deployment hook to the Activate New Release action with the pkill php5-fpm process

回答6:

Seems like the service command sometimes fails. I don't know about the reason, however I ended up using this workaround in /etc/logrotate.d/php5-fpm:

/var/log/php5-fpm.log {
    ...
    postrotate
        # The original reload command did never work
        #invoke-rc.d php5-fpm reopen-logs > /dev/null

        # Workaround for cases when the reload command fails for some reason
        service php5-fpm reload > /dev/null 2>&1
        [ $? = 0 ] || ( service php5-fpm stop; pkill php5-fpm; service php5-fpm start ) > /dev/null
    endscript
}

回答7:

The option that worked for me was to reset opcache. Create a file opcache-reset.php file in your site root and add:

<?php opcache_reset(); ?>

After reseting opcache, you can delete the file or move it out of your site root.

回答8:

When none of the above answers solves the problem, test your configuration to see your mistakes:

$ php5-fpm -t

I most cases you probably made a mistake in your /etc/php5/fpm/php-fpm.conf or /etc/php5/fpm/pool.d/www.conf

回答9:

For me it turned out to be an error in pool configuration file (loaded from pool.d directory)

Try running php-fpm7.0 -t to run configtest (name of the executable may vary based on your installation and name of the service)