PHP-FPM 5.6 and 7.0 error restart every minute

linuxpl

linuxpl

Verified User
Joined
Mar 13, 2015
Messages
23
Location
Poland
Hi,

I have big problem with php-fpm 5.6 and 7.0 - every minute php-fpm is restarted and mostly not started, after 15-20 seconds work again, but after next restart again failed.

System Debian 8.

I have second one server - this same hardware and software configuration, and everything work properly. This one is a new server - about 15 users. After rebuild them to fastcgi all works, but I want php-fpm...

Any idea?


daemon.log:
Code: 
Apr 17 00:45:06 hs2 systemd[1]: php-fpm56.service start operation timed out. Terminating.
Apr 17 00:45:06 hs2 systemd[1]: Failed to start The PHP FastCGI Process Manager.
Apr 17 00:45:06 hs2 systemd[1]: Unit php-fpm56.service entered failed state.
Apr 17 00:45:32 hs2 systemd[1]: php-fpm70.service start operation timed out. Terminating.
Apr 17 00:45:32 hs2 systemd[1]: Failed to start The PHP FastCGI Process Manager.
Apr 17 00:45:32 hs2 systemd[1]: Unit php-fpm70.service entered failed state.
Apr 17 00:45:37 hs2 systemd[1]: Starting The PHP FastCGI Process Manager...

php-fpm70.log
Code: 
[17-Apr-2017 00:41:36] NOTICE: fpm is running, pid 4750
[17-Apr-2017 00:41:36] NOTICE: ready to handle connections
[17-Apr-2017 00:43:06] NOTICE: Terminating ...
[17-Apr-2017 00:43:06] NOTICE: exiting, bye-bye!
[17-Apr-2017 00:44:01] NOTICE: fpm is running, pid 4847
[17-Apr-2017 00:44:01] NOTICE: ready to handle connections
[17-Apr-2017 00:45:32] NOTICE: Terminating ...
[17-Apr-2017 00:45:32] NOTICE: exiting, bye-bye!

journalctl -xn
Code: 
-- Logs begin at pon 2017-04-17 00:29:24 CEST, end at pon 2017-04-17 00:59:53 CEST. --
kwi 17 00:59:31 hs2.linux.pl systemd[1]: php-fpm56.service start operation timed out. Terminating.
kwi 17 00:59:31 hs2.linux.pl systemd[1]: Failed to start The PHP FastCGI Process Manager.
-- Subject: Unit php-fpm56.service has failed
-- Defined-By: systemd
-- Support: http://lists.freedesktop.org/mailman/listinfo/systemd-devel
-- 
-- Unit php-fpm56.service has failed.
-- 
-- The result is failed.
kwi 17 00:59:31 hs2.linux.pl systemd[1]: Unit php-fpm56.service entered failed state.
kwi 17 00:59:36 hs2.linux.pl systemd[1]: Starting The PHP FastCGI Process Manager...
-- Subject: Unit php-fpm56.service has begun with start-up
-- Defined-By: systemd
-- Support: http://lists.freedesktop.org/mailman/listinfo/systemd-devel
-- 
-- Unit php-fpm56.service has begun starting up.
kwi 17 00:59:48 hs2.linux.pl lfd[2008]: SYSLOG check [wDTY71znS56JJlyb4RVJXG3CEfSs]
kwi 17 00:59:53 hs2.linux.pl clamd[1329]: SelfCheck: Database status OK.
 
Last edited:
Hello,

You need to identify a reason on why PHP-FPM fails to start properly. The logs you posted does not give a clue at least to me.

You'd better read entire /var/log/messages directly without using 'journalctl -xn'.
 
I think this may be systemd related.

I changed entry Type=notify to Type=simple in file /etc/systemd/system/php-fpm70.service, and php-fpm70 now work without timeouts and restarts.

php-fpm56 work at this moment without this change, but this is copy which I made before last compiling.

But still I don't know why this work on another two servers (after last update php7.0 too).

/var/log/messages has no any entry related, there are some about segfaults php 5.6:

Apr 17 16:39:27 hs2 kernel: [58114.901479] php-fpm56[11337]: segfault at 2af889800 ip 00000000007d96fb sp 00007ffc4cfdc210 error 4 in php-fpm56[400000+a9a000]
Apr 17 16:39:28 hs2 kernel: [58115.919405] php-fpm56[11340]: segfault at ffffffff ip 00000000007d96fb sp 00007ffc4cfdc210 error 4 in php-fpm56[400000+a9a000]
Apr 17 16:39:30 hs2 kernel: [58118.436618] traps: php-fpm56[11341] general protection ip:7d9ac9 sp:7ffc4cfdc210 error:0 in php-fpm56[400000+a9a000]
 
Shooting in a dark...

do you see any error or warning when you run in root console:

Code: 
/usr/local/php56/bin/php -v

Code: 
/usr/local/php70/bin/php -v
 
No error messages.

Now I know this is not a php problem - php scripts works normal when php is in start state:
Code: 
● php-fpm70.service - The PHP FastCGI Process Manager
   Loaded: loaded (/etc/systemd/system/php-fpm70.service; enabled)
   Active: activating (start) since pon 2017-04-17 21:21:01 CEST; 1s ago
 Main PID: 29289 (php-fpm70)
   CGroup: /system.slice/php-fpm70.service
           └─29289 php-fpm: master process (/usr/local/php70/etc/php-fpm.conf)

but after about one minute php entered failed state (and php scripts stop works - error 500):
Code: 
● php-fpm70.service - The PHP FastCGI Process Manager
   Loaded: loaded (/etc/systemd/system/php-fpm70.service; enabled)
   Active: failed ([B]Result: timeout[/B]) since pon 2017-04-17 21:20:06 CEST; 53s ago
  Process: 29110 ExecStart=/usr/local/php70/sbin/php-fpm70 --nodaemonize (code=exited, status=0/SUCCESS)
 Main PID: 29110 (code=exited, status=0/SUCCESS)

kwi 17 21:20:06 hs2.linux.pl systemd[1]: php-fpm70.service start operation timed out. Terminating.
kwi 17 21:20:06 hs2.linux.pl systemd[1]: Failed to start The PHP FastCGI Process Manager.
kwi 17 21:20:06 hs2.linux.pl systemd[1]: Unit php-fpm70.service entered failed state.

I have no idea how I can diagnose more. At this time one solution for me is change Type to simple in systemd script. After this php start fast and no errors:
Code: 
● php-fpm70.service - The PHP FastCGI Process Manager
   Loaded: loaded (/etc/systemd/system/php-fpm70.service; enabled)
   Active: active (running) since pon 2017-04-17 21:22:29 CEST; 24s ago
 Main PID: 29383 (php-fpm70)
   CGroup: /system.slice/php-fpm70.service
           └─29383 php-fpm: master process (/usr/local/php70/etc/php-fpm.conf)

kwi 17 21:22:29 hs2.linux.pl systemd[1]: Started The PHP FastCGI Process Manager.
 
All Updated and Custombuild?

We had this for a about a year ago,

Then deinstalled the php 7 didn't solve the problem, placing back image and again installing php 7 problem came back, so we don't use the php 7 after that, at that time more probs.

No errors was the main problem.

PHP script don't working well if restarts, not every php programm or cms likes this. ! ;)

Problem her is/was in testing everything went good, then live there was a update wich we didn.t find out the problem starts, so while not having php 7 is here on that server not bad, we don't try to install php7 AGAIN while LIVE BOX.

CENTOS 7 PHP7 and PHP 5.6 both FPM Apache 2.4 at that time Suhossin was maybe the problem interfering while no php7 support at that time i thought as maybe the cause of this! ( also ioncube, HTTP/2 and zend installed)


So thats why i ask is all updated?

I mean all updated before putting / enable second version of PHP on the BOX

Problem was also after updates everything seems to work, but at a time after rebooting BOX this was then the error, so problem was here unoticed for a while, so debugging for real error or which update was not succesfull!

So take care of your other 2 boxes that they still work well if you reboot them? ( and or rewrite confs)

Sorry my reply is not helping you, while i only pointing out is All Updated , to reboot boxes and rewrite confs to check on other boxes to.
 
Last edited:
I think problem is solved.
The problem is that the php-fpm process doesn't return a signal to systemd when it's starting, so systemd thinks it didn't start properly and kills the process and restart them.

I addeed --with-fpm-systemd option to php70 compilation and now php-fpm70 works with Type=notify in /etc/systemd/system/php-fpm70.service!
 
Last edited:
Are you sure that you have the latest custombuild script and directadmin? It's custombuild which should care of it, I see this:

Code: 
                if ! grep -c 'with-fpm-systemd' ${CONF_FILE}; then
                        perl -pi -e 's#--enable-fpm \\#--enable-fpm \\\n\t--with-fpm-systemd \\#' ${CONF_FILE}
                fi

Or you have a custom versions of configure.php70 ?
 
Hmm... I'm sure that I have latest DA and custombuild, this server was intalled 2 months ago and all software is updated.
I see this entry in build script too, and I have custom version of configure, but diff with default is only "--enable-exif" \ - this same config work on 2 more servers.

Will try remove --with-fpm-systemd from config and build php70 again. Let's see if this works...
 
Just solved the same issue on a server of my client with PHP-FPM 7.1

In my case the client had a custom /usr/local/directadmin/custombuild/custom/fpm/configure.php71 with double quotes, i.e.

Code: 
        "--program-suffix=71" \
        "--enable-fpm" \

hence the patch in the build script did not work. Everything went fine after removing double-quotes and running:

Code: 
./build php
 
You must log in or register to reply here.
Back
Top