Phpswitcher 5.0.5 Bug - Default php pool config and sockets gets deleted when creating a new customer

  • Hello,

    since Phpswitcher plugin version 5.0.3 we had the same problem as described in PhpSwitcher 5.0.3 Problem, which got already fixed with version 5.0.4. But there was as second problem, which points at the default php version and not the ones added from the phpswitcher.

    I'll explain:

    Problem A - got resolved with 5.0.4:

    If using the apt-get upgrade the socks and pool configuration files from sites which used additional phpversions got removed and not created again. Therefore the sites failed with service 503. A simple perl /var/www/imscp/engine/imscp-rqst-mngr -v fixed this in the meantime and the problem got fixed entirely with Phpswitcher 5.0.4. But at the same time a Problem B existed.

    Problem B - exists since 5.0.3 and still in 5.0.5:

    When switching a site back to the default phpversion or when creating a new customer with default phpversion, the same problem from above now hits all pages wich are using the default php version. The command perl /var/www/imscp/engine/imscp-rqst-mngr -v is not able to fix this problem, but with a fast reconfiguration perl /var/www/imscp/engine/setup/imscp-reconfigure -danv this problem gets hotfixed aswell in the meantime. But it seems like there is still another problem hidden in the phpswitcher.

    For debugging help:

    • Create a new customer with default phpversion or switch one site back to the default phpswitcher with latest version Phpswitcher 5.0.5
    • The new customer got created successfully and the corresponding php-fpm process is restarted - in my case php7.1-fpm.
    • Only the pool configuration for the new or switched customer is created, every other page using the same phpversion, throws a service 503 error.
    • The apache2 error logs are helping - see:
    1. [Wed Oct 23 18:02:10.640520 2019] [proxy:error] [pid 27801:tid 139865844823808] (2)No such file or directory: AH02454: FCGI: attempt to connect to Unix domain socket /run/php/ ( failed
    2. [Wed Oct 23 18:02:10.640539 2019] [proxy_fcgi:error] [pid 27801:tid 139865844823808] [client REDACTED] AH01079: failed to make connection to backend: httpd-UDS, referer:
    • Using perl /var/www/imscp/engine/setup/imscp-reconfigure -danv to let i-mscp fix these problem.
    • Pool configurations and sockets for default php version gets recreated and every page is working again.

    Additional information:

    • Default PHP Version in use: PHP 7.1.32-1+0~20190902.23+debian9~1.gbp9d1be7 (cli) (built: Sep 2 2019 13:35:11) ( NTS )
    • I-MSCP Version: i-MSCP 1.5.3 Build: 2018120800 Codename: Ennio Morricone
    • OS: Debian 9.11 stretch
    • Phpswitcher Version: Current 5.0.5
    • Additional PHP version installed from package and used in phpswitcher: PHP 7.2.23-1+0~20191008.27+debian9~1.gbp021266 (cli) (built: Oct 8 2019 05:48:16) ( NTS )

    If there is any other question, I'm happy to answer and help to find the problem.

    Kind regards,


  • Nuxwin

    Added the Label to be investigated
  • badge.php?id=1239063037&bid=2518&key=1747635596&format=png&z=547451206

  • I can also confirm this.

    Default PHP Version: PHP 7.1.32-1+0~20190902.23+debian9~1.gbp9d1be7

    I-MSCP Version: i-MSCP 1.5.3 Build: 2018120800OS: Debian 9.11 stretch
    Phpswitcher Version: 5.0.5

  • I also have the problem. In the morning sometimes all pages with the standard php version do not work (for me 7.1.32). I then switch to one compiled and back again.

    i-MSCP 1.5.3, Build: 2018120800, Debian stretch, phpswitcher 5.0.5

  • Hello.

    I also have the same problem. Default php is 5.6.40 in my case.

    Is there any permanant fix for this problem?

    Cannot access that github page, is the solution is there?

  • There is an issue opened on githug. This bug will be fixed ASAP ;)


  • There is an issue opened on githug. This bug will be fixed ASAP ;)


    Could I also have access to the github repository? Or is there a new release with fix coming in near future? We really need to get this fixed ASAP. :) I also sent a private message about this on 30h Dec.