mkniskanen
Verified User
Hi,
I have updated PHP using Custombuild 2.0 but I cannot say whether the problem is in the build process or elsewhere. My problem started by getting warnings from CSF (Configserver firewall) about php-fpm processes staying alive more than half an hour. Being a moron I have neglected them. Today, however, my DA powered server (nginx+php-fpm) refused to work. In nginx error_log I found a dozen or so lines like this
Restarting php-fpm71 from DA helped. However, later on I noticed that aside from the main processess (owned by root) there were stale processes owned by the users which obviously had been spawned but not terminated. I also checked my email just to get the same warnings again.
Anybody else experienced the above? I have so far found no definitive answer or a solution. What I am running is one single instance of WordPress (yes, I know, get rid of it) for one customer and about a dozen OpenCart installations. The stale processes seem to come from both of these environments. I am not convinced that it is the PHP code that is to blame. There must be something more to this behaviour. This has been going on almost from day 0 after the installation but I have not paid any attention to it until recently.
I have to say I am very suspicious about the socket method of communication between nginx and php-fpm because this used to produce some gray hair growth earlier.
Maybe it is time for Mr root to start restarting php-fpm every six hours or so...
Edit: The CSF warning mail looks like this
Markku
I have updated PHP using Custombuild 2.0 but I cannot say whether the problem is in the build process or elsewhere. My problem started by getting warnings from CSF (Configserver firewall) about php-fpm processes staying alive more than half an hour. Being a moron I have neglected them. Today, however, my DA powered server (nginx+php-fpm) refused to work. In nginx error_log I found a dozen or so lines like this
Code:
2017/06/26 09:39:13 [alert] 12926#0: *110350 open socket #10 left in connection 4
2017/06/26 09:39:13 [alert] 12926#0: *110342 open socket #103 left in connection 7
...
2017/06/26 09:39:47 [alert] 7451#0: *9 open socket #108 left in connection 11
2017/06/26 09:39:47 [alert] 7451#0: aborting
Restarting php-fpm71 from DA helped. However, later on I noticed that aside from the main processess (owned by root) there were stale processes owned by the users which obviously had been spawned but not terminated. I also checked my email just to get the same warnings again.
Anybody else experienced the above? I have so far found no definitive answer or a solution. What I am running is one single instance of WordPress (yes, I know, get rid of it) for one customer and about a dozen OpenCart installations. The stale processes seem to come from both of these environments. I am not convinced that it is the PHP code that is to blame. There must be something more to this behaviour. This has been going on almost from day 0 after the installation but I have not paid any attention to it until recently.
I have to say I am very suspicious about the socket method of communication between nginx and php-fpm because this used to produce some gray hair growth earlier.
Maybe it is time for Mr root to start restarting php-fpm every six hours or so...
Edit: The CSF warning mail looks like this
Code:
Time: Mon Jun 26 19:52:21 2017 +0200
Account: mycust
Resource: Process Time
Exceeded: 1848 > 1800 (seconds)
Executable: /usr/local/php71/sbin/php-fpm71
Command Line: php-fpm: pool mycust
PID: 27057 (Parent PID:11830)
Killed: No
Last edited: