Recommendation - http timeouts and ERR_CONNECTION_TIMED_OUT

  • - IMSCP: 1.3.16 (Build: 20170107)
    - Distribution: Debian 8.6
    - Proftpd
    - PHP FCGID
    - MariaDB 10.0
    - Courier
    - Roundcube
    - Pydio
    - Plugins:
    PanelRedirect 1.1.5, PMA Captcha 1.1.1, RoundCubePlugins 1.2.6, SpamAssassin 1.1.1 , LetsEncrypt 3.0.0


    For many years I just used the standard mpm_worker settings.
    Now there are problems with timeouts and many ERR_CONNECTION_TIMED_OUT w/ https.


    Intel Quad-Core
    32 GB DDR3 RAM
    2x 500 GB SSD RAID1
    Domains 240
    Subdomains 108
    Domain aliases 102


    Yesterday I have changed the worker settings and it has improved slightly ... are these settings recommended with the standard fcgid settings or should I use smaller values?
    Recommendation?


    Timeout 5
    KeepAlive On
    MaxKeepAliveRequests 100
    KeepAliveTimeout 5
    HostnameLookups Off

    Edited once, last by fulltilt ().

  • I guess I need some help here ... still issues with blank pages - Monit can not acccess port 80 on localhost and restarts apache.


  • First please, update your distro and update i-MSCP to latest version. You won't receive support for older series.

    badge.php?id=1239063037&bid=2518&key=1747635596&format=png&z=547451206

  • ### Note ###
    I will not be able to do the update until the weekend ...
    But I recognized that a certain error message appears in the daemon logs before it freezes:


    Access denied for user 'web757'@'localhost' (using password: YES)


    after that I see also a DBI connection error for the imscp_user connection
    and mostly the panel is also frozen so I have to restart nginx, panel and apache.


    The system has been running smoothly for about 5 months without any problems, so it could be caused by a customer script ...
    above database connection access for web757'@'localhost') was set to % by the customer

  • @fulltilt


    Check that you have not any apache or vlogger zombie process that take all memory or CPU.

    badge.php?id=1239063037&bid=2518&key=1747635596&format=png&z=547451206

  • CPU load looks good ... no zombies and nothing special in our error logs, v3.16 was always very reliable so far ...


    for testing I have set a higher value for fs.file-max = 4097152 and reduced open-files limits in my.cnf
    there were no problems at night, but I guess it starts over when customers working on their applications ;-)


    ### edit ###
    yep, have to update - looks like the vlogger issue

    Code
    1. [Wed May 03 08:37:23.925425 2017] [mpm_worker:notice] [pid 24236:tid 140630988191616] AH00295: caught SIGTERM, shutting down
    2. [Wed May 3 08:37:23 2017] [notice] vlogger: caught TERM, shutting down
    3. [Wed May 3 08:37:25 2017] [notice] vlogger: DBI usage tracker dump interval set to 30 seconds
    4. [Wed May 3 08:37:25 2017] [notice] vlogger: started CustomLog Handler -- resuming normal operations

    Edited once, last by fulltilt ().

  • it seems sometimes there are conflicts w/ nginx and or php5-fpm, if I restart nginx and php5-fpm, apache2 is also back online ...


    and sometimes an nginx restart shows this error message:

  • @fulltilt


    Such error shouldn't occurs because Apache shouldn't listens on the 8880 and 8443 ports.


    ANYWAY PLEASE, UPDATE TO latest version. Once done, report the bug if any.


    Thread closed.

    badge.php?id=1239063037&bid=2518&key=1747635596&format=png&z=547451206