So now I have Galera MariaDB cluster working with ispcinfighting what is next

Discussion in 'Installation/Configuration' started by isn, Sep 26, 2018.

  1. isn

    isn Member

    After a great deal of struggle , I have you two node ispconfig set up working. Whether I enter on the web server IP address one old IP address to I get to see the same data and the updates propagate as expected.q

    The next step would be to set up a DNS configuration that is aware of the status of the services of the two IP addresses it services.

    As we all know what a round robin set up the same traffic to the downed IP address. Do I have to set up my own script and monitor the status of the two IP address is there a solution with bind that provides answer only if the up address is up?

    My set up is working beautifully and I’m gonna take it to the next level. I want my small service one out of google compute, two small machines.

    Although I am not if we could poster I am a frequent visitor to this website.
  2. Taleman

    Taleman Well-Known Member HowtoForge Supporter

    From what I have read browsers and some other applications try all the IP-numbers they receive. So your setup, if the round-robin works, would work for those applications already.
    I am not aware of any ready bind functions that would know when one host in the round-robin list is down, but would that help much anyway? DNS info takes hours to propagete, so by the time a dns cache on the other side of the globe gets the new DNS info, the host may well have been repaired and would be usable again.
  3. ahrasis

    ahrasis Well-Known Member HowtoForge Supporter

    I am still studying round robin but simple failover will be to set dns A record for your domain to both public ip of your web servers at TTL 180-300, as 60 - below 180 may be too fast for many ISP, though of course you can try that lower TTL.

    Do share how you intend to implement round robin in your dns server.
  4. Jesse Norell

    Jesse Norell ISPConfig Developer Staff Member ISPConfig Developer

    If you use a short ttl it could work reasonably well in theory if you don't mind a small bit of connection issues at failover time. You would have to setup a monitoring system that recognized one server was down, updated DNS, and pushed a new zone file. That's about the same amount of work as having a monitor catch the outage and failover the IP address, which should be faster, with less connection problems.

    Another issue with DNS round robin balancing is that the other end doesn't always expire old records and/or reconnect, even when you have the ttl short and update the DNS zone promptly. We used to have DNS round Robin setup for a mail server, and clients connected to the down server would often stay down until they restart their email program (sometimes even their computer, or clear the DNS cache). This issue of course depends on the OS used and the mail client in use, but it was a enough of a problem that we went away from that design. That has been a few years, though, and things could be in better shape now.

    One option that can be easy to setup if your firewall/router supports it is to load balance right there. That particular device already has to be up, so less of an "one more place to fail" issue. If you don't have that ability, you could use an external load balancer, either an appliance or even just a light weight server running haproxy or similar.

    In this particular case, I would first investigate what failover options are available right in the Google platform, since that's where it's hosted. I have no experience there, and barely remember what I've read on it in the past, so no recommendations, but might be worth looking at.

Share This Page