I run Windows NLB on 2 cf7 enterprise servers each with 2 clustered instances of CF - to be honest, I've never had any problem whatsoever. Uptime of the servers was up till recently just shy of 600 days (and that's on Windows!!!! - and don't go on about not applying patches etc) CF has been rock solid, never missing a beat, I've never had to rely on a single CF instance on one box and the load balancing whilst primitive works fine for what we need. It's not application level aware so when the SAN hosting the web files goes south then the file not found message isn't caught but use trusted cache and that problem is alleviated. We're running none sticky sessions, the two nodes are in their own VLAN across 2 data centres - windows NLB causes significant broadcast storms on all ports of a switch and if you've got a lot of boxes in the same switch they'll all receive the stormed packet!
as for only just shy of 600 days - an odd thing happened the other day, all our HP servers of a certain model restarted at 6am the other day with no warning or anything in the event log!!! We had none web servers in the list of affected boxes too - no power outages nothing - just can't explain it! hope that helps, john. On 1/31/07, Jochem van Dieten <[EMAIL PROTECTED]> wrote: > WebSite CFtalk wrote: > > The problem is not always there, but happens especially with one > > customer using clustered terminal servers and Internet Explorer. > > > > It seems that they are always connected to one of the cluster nodes > > Are the terminal servers in the same subnet as your cluster? If so, what is > the ARP timeout on the system? If not, what is the ARP timeout on the > upstream router of your cluster? Do you have port security enabled on your > network equipment? Are you using multicast and / or IGMP? How is multicast > configured on your switch / router? Do you have single affinity set? > > > > they experience drops regularly, also when all jrun/cf instances is up > > and running. > > I would look into the combination of NLB with your network equipment. ISTM > this is a problem where the network directs the traffic to one server, where > NLB tries to direct the traffic to another server. > > > > We're shortly going to upgrade to 7.x (running 6.1 right now), but I > > believe that the problem is with NLB clustering - maybe when the other > > end is also NLB. > > > > The plan is to switch to a HW based solution for this reason and also > > since it obviously would be a better and more stable/powerful solution. > > This all sounds very vague to me. Wouldn't it be better to thoroughly > investigate the issue before investing in new hardware? Or at least get > something from a vendor on a try before you buy basis. > > Jochem > > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~| Upgrade to Adobe ColdFusion MX7 Experience Flex 2 & MX7 integration & create powerful cross-platform RIAs http:http://ad.doubleclick.net/clk;56760587;14748456;a?http://www.adobe.com/products/coldfusion/flex2/?sdid=LVNU Archive: http://www.houseoffusion.com/groups/CF-Talk/message.cfm/messageid:268179 Subscription: http://www.houseoffusion.com/groups/CF-Talk/subscribe.cfm Unsubscribe: http://www.houseoffusion.com/cf_lists/unsubscribe.cfm?user=89.70.4
