HHVM is acting up

Discussion in 'General' started by Geekteq, May 24, 2017.

  1. Geekteq

    Geekteq New Member

    Hey, been googling on this forever so now I gave up and here goes.

    I installed HHVM with the 3.1 update and it's listed as selectable and it also starts when I assign it to a host.
    Works brilliantly. But...

    About 4-6 hours after the host has started the HHVM-instance for the host goes away and doesn't come back unless I do a restart of that hosts HHVM.

    I can't find any reason for it.. The only thing the host error log says is:
    [Wed May 24 19:15:43.126629 2017] [fastcgi:error] [pid 10293] (111)Connection refused: [client] FastCGI: failed to connect to server "/var/www/clients/client3/web9/cgi-bin/hhvm-fcgi-*-80-hostname.tld": connect() failed

    Which states what I already know.. the hhvm-instance has gone away -- I've tried enabling more logging for HHVM in /etc/hhvm/* but my error log is empty..
  2. Jesse Norell

    Jesse Norell ISPConfig Developer Staff Member ISPConfig Developer

    Run 'dmesg' and see if there is anything (though those should be in one of your log files, too). You could attach to the hhvm process with strace, logging output to a file, and wait for it to die again to see what happens.

    In my experiments I found hhvm (from debian jessie) to be a huge memory hog (at least for websites running the wordfence firewall), and after many times of increasing memory settings I just gave up on it. I have picked up the indication from other comments that it is just buggy at times, so you may be hitting a bug that crashes it. (I never dug further into the memory bloat because they want you to compile and test the latest version - which is a reasonable position to take, but it wasn't worth the effort to me).

Share This Page