Re: Life getting to complicated?

Mark Rousell <mark.rousell@xxxxxxxxxxxxx> · Wed, 24 Apr 2019 02:03:26 +0100



    On 23/04/2019 22:24, Lester Caine wrote:

    This was an attempt to rule out the Hub6 as the
      problem, but with the Netgear ALSO dropping the connection not
      sure quite where we are ...
      

    Sounds like it must be in BT's or Openreach's equipment. Either
    something wrong with the copper between you and the cabinet, a port
    on the equipment in the cabinet, or maybe something at the exchange.

    
      The main problem here is that the 'problem' sites are ones running
      Wordpress. My own sites don't have a problem as things like cron
      happen independent of PHP, but it was the Wordpress scheduled
      events that were messed up ... along with email but that turned
      out to be just a corrupt IP change ...
      

      In both cases however one accesses the dashboard via the domain
      name in order to get the at the management toolsc.
      

    If you access the Wordpress dashboard via domain name then you could
    almost certainly access it via private LAN IP address. If you need
    host header to be able to reach the dashboard website on the server
    then either a LAN DNS server or edited hosts file (as suggested by
    Curtis Maurand) would allow you to do it.

    
    As Curtis suggested, if it is too fiddly to change the
    script/Wordpress configuration to use local LAN IP addresses then
    editing your hosts file would work. Entries similar to

    
    192.168.1.33    dashboard.myserver.com

    
    would do it. Machines with a hosts file entry like this would skip
    DNS lookups and be given the internal LAN IP instead.

    
    Alternatively you could run an internal cacheing DNS server. In this
    scenario, the local DNS server would contain entries for local
    servers which would resolve to local LAN IP addresses, and would
    pass on any other DNS query for the rest of the world to an upstream
    DNS server. For the avoidance of doubt, the local DNS server would
    be for internal LAN use only. You could use Pi-hole for this (I
    think), PFSense (as Curtis suggested) or OPNSense, or almost any
    other DNS server software of your choice.

    
      There are four local servers but only 2 are visible from outside.
      One on each broadband line, both of which have static IPV4
      addresses (but only BT support static IPV6 - Vodafone asked what
      it was when I requested it)
      

      The servers are SUSE Linux, Nginx, assorted PHP-FPM for 5.x, 7.0
      and 7.2 along with Firebird and MySQL for the Wordpress sites.
      Obviously in an ideal world all four servers would be visible from
      either line so that the BT one going down would not be a problem
      at all. Just reduced bandwidth till it's back 10 minutes later ...
      and I THINK the starting point for that is to have my own DNS
      servers on each line but now I am getting out of the comfort zone
      ... what I have does work the vast majority of the time so perhaps
      I leave it like that :)
      

    This can get as complicated (and expensive) as you want. :-) Also
    it's not a PHP issue as such but, until someone tells me off, I'll
    comment here.

    
    I think you could do this with round robin DNS. You could have one
    public-facing DNS server on each broadband line (or even the same
    server serving both lines). The server or servers would be the
    authoritative DNS server(s) for the domains in question. These
    servers would resolve public DNS queries for the web servers in a
    round robin fashion, distributing served IP addresses roughly
    equally between the two lines. Ideally you'd need some sort of
    heartbeat checking so that the DNS server(s) can detect when one of
    the lines or servers has gone down and remove that line's or
    server's public IP DNS entry from DNS responses being served out
    (and replace it when the server or line comes back up).

    
    -- 
Mark Rousell