On 12/7/05, Rick Root <[EMAIL PROTECTED]> wrote: > I'm getting suspicious because this is the third message I've sent to > the list in the last few weeks that generated no responses whatsoever. > Doesn't anyone have any comments on failover/redundancy in a coldfusion > environment? >
Rick, We run a clustered environment where we required (at least hoped for) both failover and load balancing. We are currently using session failover (at the JRun level) and have a BigIP hardware load balancer in front of our two physical boxes serving requests to each. While the BigIP load balancer is great, session failover at the JRun level has been (and still is) a pain. I have sticky sessions on at both the JRun level (in the jrun.xml file) and at the BigIP level, yet I still get errors saying that memory structures that I know *have* to exist based on the page that failed don't exist. Here's some things to consider when making your decisions: 1) Remember that CFCs are (frustratingly) not able to be replicated by JRun, so you will need to cobble together a workaround if you intend to support session failover (i.e., storing the instances variables in the session scope as a ColdFusion structure and then re-creating the object on the target page, passing in the instance variables to the constructor). 2) Make sure you're on the latest JRun Updater, which is 6. Clustering, in my opinion, was just flat-out broken in the first couple of releases, but starting with Updater 4, Macromedia (or is it Adobe now?) got their act together and fixed up many of these issues. If you're on MX 7.0.1 using the intregrated JRun, you should be fine here. 3) Be careful with session replication, if this is a feature you want. I honestly don't trust it. That's sort of a general statement I understand, but this is about 3 years of dealing with this telling you this, not somebody who just played around with it as an experiment. 4) Hardware load balancers are very expensive, though you can usually find good deals on e-Bay if you really want one. I love our BigIP balancer, but it's certainly not necessary for all environments. I'm not too familiar with the software-based solutions, so I can't help you there. 5) Stay away from NAS devices for central repositories. I've read in several places that these devices can become severe bottlenecks in a fast-request environment such as the web. We're in a *nix setup and simply have a NFS mount that our boxes share. It works perfectly fine and there's no need for additional hardware. 6) Test, Test, Test. Make sure you run tests where your boxes are both functioning and then shut one down. Make sure requests are seamlessly served by the other box. Specifically you want to make sure that, if Box A serves a request that sets session variables and you shut it down, Box B will pick up those changes without issue. Other people's experiences will certainly vary from mine. Setting up a full-featured clustering environment is not for the faint of heart, and I certainly advise you to read up and map out exactly what you want to do before hand. The negative tone that this post might imply is not intended to scare you off by any means. But it's not a job that only takes an hour or two either. Hope this helps? Regards, Dave. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~| Discover CFTicket - The leading ColdFusion Help Desk and Trouble Ticket application http://www.houseoffusion.com/banners/view.cfm?bannerid=48 Message: http://www.houseoffusion.com/lists.cfm/link=i:4:226428 Archives: http://www.houseoffusion.com/cf_lists/threads.cfm/4 Subscription: http://www.houseoffusion.com/lists.cfm/link=s:4 Unsubscribe: http://www.houseoffusion.com/cf_lists/unsubscribe.cfm?user=11502.10531.4 Donations & Support: http://www.houseoffusion.com/tiny.cfm/54
