Skip to main content
F
Fate
Participant
June 26, 2007
Question

Moving single server to HA

  • June 26, 2007
  • 1 reply
  • 331 views
Hey all,

I currently have a CF site I wrote some time back as a bill payment system. It is running on CF 6.1.0 currently.
We have new directive to make this site HA, so along with an upgrade to CF 7+ we would also like to put a second server in place for full HA compliance. I have previously set up a server for another project using the built in clustering feature with CF 7 enterprise and have not been impressed with its overall functionality. Regardless this new project requires full server hardware failover. I have been scouring google for some time trying to determine if CF 7 can be cluster aware, and if so, how do I go about implementing it? I think CF used to be able to do full hardware failover with ClusterCATS, but this is not so anymore.
Do I need to also budget in a Network Load Balancing piece also in order to make this happen, or is there a way to do this without the additional hardware? If I do buy an NLB, will CF on both servers be able to talk to eachother (are they cluster aware)? It is very important that in the case of a server failure that traffic redirects to the active server and that it maintains session state throughout....

Any help would be great as I need to report to the head guy really soon with a plan...

Thanx.
    This topic has been closed for replies.

    1 reply

    S
    Steve Sommers
    Legend
    June 26, 2007
    I'm not sure what HA is but for us us, we tackled the clustering question as follows:

    First, we spec'd out an "ultimate dream" system. No idea (no matter how stupid ;-) was rejected. Next, we spec'd out our "minimum requirements" system. This was probably the hardest part because different people have different opinions on "must have" vs. "nice to have." Then we determined development ETA's and estimated all costs, both in time and dollars. After putting all this together, we opted for our "ultimate dream" system with pretty much only one short-coming: clustering.

    We provide a "server farm" to the point that once a user logs on, the entire session is hosted on a single server. For us, the costs associated with true clustering far outweighed the minimal benefits that would have been gained. With the alternative we came up with, worse case, if a server blows up sending shrapnel everywhere, users would have to relogin, and this process would bounce them to a new server -- no big deal. Now if servers regularly blew up, then we might have come to a different conclusion.

    You might want to rethink wht you would really gain from true clustering vs. other far less costly alternatives.
    Terms & ConditionsAccessibility statement
    Using the community Terms of use Privacy Cookie preferences Do not sell or share my personal information ad choicesAdChoices