Nginx PHP-FPM Nginx keeps crashing?

Centmin Mod is provide as is, so short of scripted related bugs or issues, any further optimisation to the web stack components - nginx, php-fpm, mariadb mysql, csf firewall etc or web app specific configurations are left to the Centmin Mod user to deal with. So I do not provide any free support for such.

Nginx 502 or 504 Bad Gateway Errors

Bad gateway 502 /504 timeouts are usually related to Nginx timing out waiting on PHP-FPM to respond as PHP-FPM is overloaded or overwhelmed with requests, so may need to tune PHP-FPM values. It also maybe due to PHP-FPM in turn being queued and backed up waiting on MariaDB MySQL server to respond - so also need to look at MySQL.

You'll need to tune your PHP-FPM settings with php-fpm main pool config file at /usr/local/etc/php-fpm.conf (overview of config files) and this is left up to end user to do but here's a thread for starters to enable php status page output outlined at

• http://centminmod.com/phpfpm.html#browserstatus and
• https://community.centminmod.com/threads/pm-max_children.479/ and
• https://community.centminmod.com/threads/warning-pool-www-server-reached-max_children-setting-50-consider-raising-it.791/ which outlines the official PHP-FPM config documentation as well.

Checking PHP-FPM etc logs

You'll also need to check into your PHP-FPM, Nginx and MariaDB logs which you can find as outlined at How to troubleshoot Centmin Mod initial install issues

Server logs include Nginx, PHP-FPM, MariaDB MySQL error logs as well as others. You can find your Centmin Mod install/menu logs at FAQ 7 and server logs at FAQ 19 at Centmin Mod FAQ (most up to date info in FAQ so always read that first). Spoiler tag below has info too but may not be up to date.

Some of Centmin Mod's installed software will have their own access and error logs which maybe useful for diagnosing errors or give info, notes, or warning notices.

Note: There's no support provided by me for diagnosing such errors which may occur for various reasons including misconfiguration of installed php/mysql scripts or applications.

In SSH2 telnet you can use tail command to view the last X number of lines in the file.

For example for viewing last 10 lines in the file for:

For Nginx access and error logs:

Code:

  tail -10 /usr/local/nginx/logs/access.log
  tail -10 /usr/local/nginx/logs/error.log

For specific domainname.com access and error log:

Code:

  tail -10 /home/nginx/domains/domainname.com/log/access.log
  tail -10 /home/nginx/domains/domainname.com/log/error.log

For other system error logs located at /var/log:

list /var/log files in ascending time order so the most recently modified files are at the bottom

Code:

  ls -lhrt /var/log

Code:

total 2.7M
-rw------- 1 root  root    0 Aug 29 15:33 tallylog
-rw------- 1 root  root    0 Aug 29 15:33 spooler
drwx------ 3 root  root 4.0K Aug 29 15:35 samba
drwxr-xr-x 2 root  root 4.0K Aug 29 15:35 mail
-rw-r--r-- 1 root  500     0 Oct  8 18:13 dmesg.old
-rw------- 1 root  500     0 Oct  8 18:13 boot.log
-rw-r--r-- 1 root  500     0 Oct  8 18:14 dmesg
drwx------ 2 root  root 4.0K Oct  8 18:14 httpd
drwxr-xr-x 2 root  root 4.0K Oct  8 19:08 php-fpm
-rw-rw---- 1 mysql root 2.3K Oct  9 12:38 mysqld.log
-rw------- 1 root  root 9.2K Oct 26 10:48 yum.log
-rw------- 1 root  utmp  94K Nov  7 22:59 btmp
drwxr-xr-x 2 root  root 4.0K Nov  8 00:00 sa
-rw------- 1 root  root 269K Nov  8 21:39 messages
-rw------- 1 root  root 110K Nov  8 23:08 secure
-rw-rw-r-- 1 root  utmp  43K Nov  8 23:08 wtmp
-rw-r--r-- 1 root  root 144K Nov  8 23:08 lastlog
-rw------- 1 root  root  69K Nov  8 23:08 lfd.log
-rw------- 1 root  root 332K Nov  8 23:08 maillog
-rw------- 1 root  500  1.6M Nov  8 23:10 cron

For PHP-FPM error log:

Code:

  tail -10 /var/log/php-fpm/www-error.log

and/or

Code:

  /var/log/php-fpm/www-php.error.log

For MySQL / MariaDB error log:

Code:

  tail -10 /var/log/mysqld.log

For CSF firewall LFD log:

Code:

  tail -10 /var/log/lfd.log

For Mail log:

Code:

  tail -10 /var/log/maillog

For Cron job logs:

Code:

  tail -10 /var/log/cron

How to edit php.ini and php-fpm configuration files ?

Centmin Mod install created command short cuts outlined here to allow you to quickly edit your /usr/local/lib/php.ini file and your /usr/local/etc/php-fpm.conf file. Full list of command shortcuts below:

• Edit php.ini = phpedit ( /usr/local/lib/php.ini )
• Edit my.cnf = mycnf ( /etc/my.cnf )
• Edit php-fpm.conf = fpmconf ( /usr/local/etc/php-fpm.conf )
• Edit nginx.conf = nginxconf ( /usr/local/nginx/conf/nginx.conf )
• Edit (nginx) virtual.conf = vhostconf - only edits /usr/local/nginx/conf/conf.d/virtual.conf not the additional vhost domain.com.conf files added later
• Edit (nginx) php.conf = phpinc ( /usr/local/nginx/conf/php.conf )
• Edit (nginx) drop.conf = dropinc ( /usr/local/nginx/conf/drop.conf )
• Edit (nginx) staticfiles.conf = statfilesinc ( /usr/local/nginx/conf/staticfiles.conf )
• nginx stop/start/restart = ngxstop/ngxstart/ngxrestart
• php-fpm stop/start/restart = fpmstop/fpmstart/fpmrestart
• mysql stop/start/restart = mysqlstop/mysqlstart/mysqlrestart
• nginx + php-fpm stop/start/restart = npstop/npstart/nprestart
• memcached stop/start/restart =memcachedstop/memcachedstart/memcachedrestart
• csf stop/start/restart = csfstop/csfstart/csfrestart

Troubleshooting Tools

However, there's many linux tools and scripts that can help you figure out what was causing the load issues and when.

Tools and commands you will want to read up on and learn for basic system admin tasks and troubleshooting.

• 20 Command Line Tools to Monitor Linux Performance
• 13 Linux Performance Monitoring Tools - Part 2
• 80 Linux Monitoring Tools for SysAdmins - Server Density Blog
• Amazing ! 25 Linux Performance Monitoring Tools
• 20 Linux System Monitoring Tools Every SysAdmin Should Know
• How to monitor disk I/O in Linux from command line - Xmodulo
  
• 10 Useful Sar (Sysstat) Examples for UNIX / Linux Performance Monitoring
  
• Generate CPU, Memory & I/O report using SAR command
  
• My mysqlmymonlite.sh script can help with that mysqlmymonlite.sh README text file has instructions for setting up a cronjob to email you reports i.e. every 10/15min or every hour.

Notes:

• Centmin Mod already installed Glances tool. It can be invoked from SSH command glances or top2. Official Glances site at nicolargo.github.io/glances/ and documentation here.

However, Centmin Mod users are free to help each other out and ask questions or give answers on this community forum. My hopes are that this community forum evolves so that more veteran long time Centmin Mod users help new Centmin Mod users out

 

That unix socket is for phpmyadmin.sh addon installed phpmyadmin install with it's own separate php-fpm pool. It would happen when you try to start php-fpm service when there is already a running php-fpm service so might not be related to your 502 bad gateway errors

example running php-fpm service

Code (Text):

service php-fpm status    
php-fpm (pid 13547) is running...

Code (Text):

ps aufxw | grep php-fpm | grep -v grep
root     13547  0.0  0.4 534204  9092 ?        Ss   04:56   0:00 php-fpm: master process (/usr/local/etc/php-fpm.conf)

now try starting php-fpm service when it's already running

Code (Text):

service php-fpm start  
Starting php-fpm [02-Jun-2017 04:59:24] ERROR: An another FPM instance seems to already listen on /tmp/phpfpm_myadmin.sock
[02-Jun-2017 04:59:24] ERROR: FPM initialization failed
 failed

proper restart

Code (Text):

service php-fpm restart
Gracefully shutting down php-fpm . done
Starting php-fpm  done

or proper stop and start

Code (Text):

service php-fpm stop
Gracefully shutting down php-fpm . done

service php-fpm start
Starting php-fpm  done

Code (Text):

tail -100 /var/log/php-fpm/www-error.log

[02-Jun-2017 04:56:37] ERROR: An another FPM instance seems to already listen on /tmp/phpfpm_myadmin.sock
[02-Jun-2017 04:56:37] ERROR: FPM initialization failed
[02-Jun-2017 04:59:24] ERROR: An another FPM instance seems to already listen on /tmp/phpfpm_myadmin.sock
[02-Jun-2017 04:59:24] ERROR: FPM initialization failed

 

Timestamps look the same as previously posted ones so any cronjobs running around that time? Check other logs around that time too

what's output for cronjob listing command

Code (Text):

crontab -l

could be wordpress auto installer's auto update cronjob random sleep formula at play for centmin.sh menu option 22 based wordpress installs where you have more than one wordpress install and the wp auto update cronjob's sleep random 3 digit seconds is very close together. If it is, then you can just change the sleep interval to not be so close together. Or change the cronjob schdule completely.

also added fix for 123.09beta01 for new centmin.sh menu option 22 installed wordpress instances Beta Branch - update inc/wpsetup.inc cronjob sleep time fix

 

yeah can pm me or to check just wp updater related, just filter for sleep entries using command

Code (Text):

crontab -l | grep wp_updater

i.e.

Code (Text):

crontab -l | grep wp_updater
0 */8 * * * sleep 268s ;/root/tools/wp_updater_domain.com.sh >/dev/null 2>&1

 

for privacy, you can post results for command

Code (Text):

crontab -l | grep wp_updater | awk '{print $1,$2,$3,$4,$5,$6,$7}'

which only prints first 7 fields up to sleep XXXs

 

That's ALOT of wordpress installs there ! Can see why there'd be overlap for wordpress auto updating ! The fix I made, checks that the first 2 characters for each XXXs sleep time so make sure the inserted new cronjob doesn't match any of existing cronjob's first 2 characters. So sleep 326s and sleep 3202s or 108s and 109s won't be possible on updated installs as first 2 characters of 32 and 10 would match.

Guess I might need to randomise the /8 (every 8hrs) too. For now manual fix would be just to manually change the cronjob schedules so they don't overlap.

 

yes make it unique but changing /8 might not be best as /7 means every 7hrs, /4 means every 4 hrs which would make matters worse as it would be too frequent.

For coding, need to come up with a better way as clearly this wouldn't scale for dozens of wordpress installs heh

 

ah increasing the value should be fine though it would mean some wp instances don't get updated every 24hrs i.e. /30 would = every 30 hrs

 

Actually incrementing still has chance of overlap i.e. /6 and /8 so every 6 and 8 hrs would overlap at 24th hour mark

It's nginx + php-fpm restart using nprestart command to clear wp caching and zend opcache caching after wp updates.

though for you you could incremant by 1hr each entry - this command would print out the new cronjob listing for wp_updater related commands

Code (Text):

crontab -l | awk '/wp_updater/' | xargs -n10 | while read m hr a b c d e f g h; do i=$(($i+1)) ;HR=$(echo $hr| sed -e 's|\*\/||'); HRR=$(($HR+$i)); echo "$m "*/"$HRR $a $b $c $d $e $f $g $h"; done

for folks looking on, public snipped for privacy version would be like

Code (Text):

0 */9 * * * sleep 3202s
0 */10 * * * sleep 109s
0 */11 * * * sleep 316s
0 */12 * * * sleep 204s
0 */13 * * * sleep 728s
0 */14 * * * sleep 220s
0 */15 * * * sleep 330s
0 */16 * * * sleep 398s
0 */17 * * * sleep 365s
0 */18 * * * sleep 269s
0 */19 * * * sleep 272s
0 */20 * * * sleep 167s
0 */21 * * * sleep 108s
0 */22 * * * sleep 326s
0 */23 * * * sleep 301s
0 */24 * * * sleep 401s
0 */25 * * * sleep 145s

of course the command you run would have full cronjob entries included

that should suffice for your situation, though need to come up with better way in 123.09beta01

 

instead of manually make changes, you can use some SSH commands to do it by backing up existing crontab list to file named cronjoblist, the using cronjoblist to build the new cronjobs in new-cronjoblist file then using sed to remove all lines containing the word wp_updater. Then concatenate the new-cronjoblist generated lines into cronjoblist backup file and then setup as cronjob overwriting existing cronjob list

Code (Text):

crontab -l > cronjoblist
touch new-cronjoblist
cat cronjoblist | awk '/wp_updater/' | xargs -n10 | while read m hr a b c d e f g h; do i=$(($i+1)) ;HR=$(echo $hr| sed -e 's|\*\/||'); HRR=$(($HR+$i)); echo "$m "*/"$HRR $a $b $c $d $e $f $g $h"; done >> new-cronjoblist
sed -i '/wp_updater/d' cronjoblist
cat new-cronjoblist >> cronjoblist
crontab cronjoblist
rm -rf cronjoblist
crontab -l

might look into revising wp_updater to use flock

• Elegant locking of BASH program - Say what?
• Follow-Up: Bash script locking with flock
• using flock to protect critical sections in shell scripts

or something similar

 

Thinking of using flock for wp_updater's nprestart command. Did a quick experimental test creating 3 scripts restartphp1.sh, restartphp2.sh and restartphp3.sh with following code to just run nprestart command. If the locked file is still locked, report already detected running process

Code (Text):

#!/bin/bash
{
exec 99>/tmp/wp_updater_phpfpm.lock
if flock -n -x 99; then
/usr/bin/nprestart
rm -rf /tmp/wp_updater_phpfpm.lock
else
echo "already detected running process"
fi
echo
echo "---"
echo done
echo "---"
} 2>&1 | tee "$0.log"

setup the 3 files to run at exact same scheduled 18:02 times

Code (Text):

02 18 * * * /root/tools/cronjob/restartphp1.sh >/dev/null 2>&1
02 18 * * * /root/tools/cronjob/restartphp2.sh >/dev/null 2>&1
02 18 * * * /root/tools/cronjob/restartphp3.sh >/dev/null 2>&1

so they ran at same time

Code (Text):

tail -l /var/log/cron | grep restart
Jun 29 18:02:01 centos7 CROND[9580]: (root) CMD (/root/tools/cronjob/restartphp2.sh >/dev/null 2>&1)
Jun 29 18:02:01 centos7 CROND[9581]: (root) CMD (/root/tools/cronjob/restartphp1.sh >/dev/null 2>&1)
Jun 29 18:02:01 centos7 CROND[9582]: (root) CMD (/root/tools/cronjob/restartphp3.sh >/dev/null 2>&1)

Each script logged to a file named same as script

Code (Text):

cat restartphp2.sh.log 
Restarting nginx (via systemctl):  [  OK  ]
Gracefully shutting down php-fpm . done
Starting php-fpm  done

---
done
---

Code (Text):

cat restartphp1.sh.log             
already detected running process

---
done
---

Code (Text):

 cat restartphp3.sh.log 
already detected running process

---
done
---

Double check /var/log/php-fpm/www-error.log php log file and it's clear of errors for Address already in use / unable to bind listening socket for address messages.

So essentially nprestart command was skipped for restartphp1.sh and restartphp3.sh as restartphp2.sh had exclusive lock on /tmp/wp_updater_phpfpm.lock. Guess could of just used normal lockfile too.