Author Topic: Soldat Cluster Failover?  (Read 810 times)

0 Members and 1 Guest are viewing this topic.

Offline PureGrain

  • Major(1)
  • Posts: 10
Soldat Cluster Failover?
« on: August 19, 2007, 09:10:05 pm »
I have managed to get soldat to failover in a 2node cluster. It fails over without a hitch and fails back without a hitch, however, it manually has to be failed back right now. I will post more on this as I get going, just built the cluster today and only had about 2 hours of testing. It is public ato be tested.

I know it is a bit of over-kill but it was fun getting it too work.

Specs:
Main Server: RHEL 4.4/VMWare - Intel Celron 1.10
¬  ¬  ¬  ¬ Primary Node: VMWare RHEL 4.4 "Soldat runs here"
       Secondary Node: VMWare RHEL 4.4 "Soldat runs here"
NAS Server - FreeNAS - Intel Celron 1.10 "Soldat server files here"
DNS Server - CentOS4.4 - AMD Athlon2400+

Here is the skinny.

Soldat runs on both servers at the same time sharing the same server files on the NAS server and binds to ONE IP which is the primary node IP. If the primary node fizzles out for more than 5 seconds it will fail-over to the secondary almost seemlessly and it takes over the IP and keeps soldat alive. If the server yet again flakes out on the secondary node it will fail-back to the primary node the same way.

Hard testing has not been done on this as I just got it up and running in the last 2 days. I will post more as I get info. And yes, I plan on writing up a NICE long tutorial on how to do this when I am finished.


Offline xReaperx

  • Major(1)
  • Posts: 45
Re: Soldat Cluster Failover?
« Reply #1 on: September 14, 2007, 01:27:45 am »
That sounds pretty nice. So all up i am assuming you need 3 servers?

Offline PureGrain

  • Major(1)
  • Posts: 10
Re: Soldat Cluster Failover?
« Reply #2 on: September 14, 2007, 09:32:07 pm »
Well, you can do the NAS and DNS servers in one server. But the idea of redundancy (hence fail over) you would run the game servers on two separate servers or two separate VM/VPS/VE. The NAS is used for storage of the files that is needed by the cluster. I still have this active but have not spent much time on it lately. This is fairly smooth in it's current setup but there was an issue with fail back manually. I will post more soon when I get time to tinker again. Thanks for the interest in it. :)

Offline FliesLikeABrick

  • Administrator
  • Flamebow Warrior
  • *****
  • Posts: 6142
    • Ultimate 13 Soldat
Re: Soldat Cluster Failover?
« Reply #3 on: September 14, 2007, 11:30:27 pm »
If the two machines are on the same network, it would be better to do the failover with the IP, not DNS (that way you don't need 1s DNS TTLs and such).

Just use linux-ha (heartbeat) to watch the services, and if one goes down, have the other bring up the IP of the other machine and continue on.  All of your other concepts with the shared filesystem will work there too.

Good work.

Offline PureGrain

  • Major(1)
  • Posts: 10
Re: Soldat Cluster Failover?
« Reply #4 on: September 16, 2007, 08:50:45 pm »
Very true and in fact, I am not even using the DNS server. I am using the fail over by IP at the moment. I did forget to mention I had to setup Soldat to run as a service to get this to work. Simple init.d script. Nothing fancy. I plan on (when time gives) to start this process over from scratch and document the steps for those who want to play with soldat/clustering. I have a new server up in a local DC for Dev/Test and will be setting up some of these for public use soon. I want to test a few ideas, not only for the cluster but some other things I am working on as well. Thanks for the comments.