Forum Discussion
NimbostratusMore than 2 devices in Active-Active mode?
As for v11.4 capabilities here are the release notes that talk to that feature
http://support.f5.com/kb/en-us/products/big-ip_ltm/releasenotes/product/relnote-ltm-11-4-0.htmlrn_new
"Connection mirroring for device groups: Prior to this release, you could only implement connection mirroring between a static pair of BIG-IP devices. Now, connection mirroring is based on traffic groups, so that an active traffic group can mirror its connections to its standby peer on the next-active device, regardless of which device in the device group is the next-active device"
Take a look at the user guide at
http://support.f5.com/kb/en-us/products/big-ip_ltm/manuals/product/bigip-device-service-clustering-11-4-0/9.html
CirrostratusActive active is a total pain in the butt to manage and I would avoid it at all costs. It sounds great on paper and to executives, but the daily operation of it would add allot of work to your day.
It would be up to you to manually balance out the load for every virtual address by putting them in different traffic groups. Also, depending on the number of devices you have and the normal load percentage on them, the number of traffic groups might have to be double or more the number of LTMs you have. This is because you may need to split the load from one failed LTM across multiple devices that are in standby for those traffic groups.
Let's say you have 3 devices running at 55% load each. If one LTM fails then another device will pick up the traffic group and that device will be running at 110%. So in that case you need 6 traffic groups so that a failure of one device would split the load of two traffic groups (27.5% each) across the remaining devices so that each would have 82.5%.
Not a whole lot of fun in my opinion.
- Well said. Even though there is some scenarios where it makes sense, I would never recommend active unless it's really needed, and even then with ample redundancy in terms of performance.
- Hamish
Cirrocumulus
I never buy this argument about 'overloading' on failure of a cluster node. Good planning can alleviate that. As can keeping an eye on capacity. Plus running active/active generally means less downtime in the event of a failure, and fewer surprises when you discover that your standby system isn't quite as ready to take load as you thought... H - Eric_Flores_131"Good planning can alleviate that. " That is kind of the point I was making. All of that "planning" is the manual distribution of VIPs and traffic groups.
