BIGIP LTM Ver 9.4.5 Active/Standby Different Data Centres

Yes it can be done. I've done data centres up to 15km apart...

 

 

The things to remember are

 

 

1. For this distance, the best way would be via GTM. Except you need to have an end-2-end DNS infrastructure before it becomes practical. Some orgs don't still today.

 

 

Otherwise...

 

 

2. The front-end VLAN needs to be spanned. Because VS IP's don't change their subnets when failing over from A to B.

 

3. If the F5's don't have direct access to the back-end server VLAN's you need to do SNAT. Which works fine, except it is always a pain to debug, and some apps just do care about the client IP address.

 

4. You really want to ensure you have a non-network heartbeat link setup. I usually use either a dark fibre. e.g. a 1Gb LX SFP optical driving a reasonably shorrt fibre. But you're too far apart for that (LX will do about 10km, and the F5's don't support longer distance opticals). So you need something like a pair of switches for the heartbeat with opticals that can do 80km. e.g. 3750E's with DWDM opticals (If you have a DWDM link). Remember the heartbeat carries mirror traffic so a piece of wet string isn't good enough. It needs to have the bandwidth to carry ALL your mirrored traffic. Best bet is to have it the same speed as your front-end VLAN. Then there's no worries about it.

 

 

 

Apart from that, the distance is nothing.

 

 

H

 

Thanks for the info.

 

Yes front-end vlan is spanned.

 

Yes we do use SNAT.

 

We have 2 6509s in which fibre terminate. From there on heartbeat vlan is spanned between 2 switches. We have made it seperate, only to be used as Heart beat.

 

Bare in mind if that heartbeat link fails, even for a few seconds you will have a far far bigger mess than you would have had with standalone units.

 

We had a switchboard failiure on one of the 6509's we were heatbeating over, it reset and recovered about 12 seconds later, but the damage was done, unit 1 and 2 were active and ARP didn't sort itself out, leaving traffic going into A, through the datacentre and back to B, which had no persistance enabled and dropped the traffic.

 

Mac masqurading might have helped, but we solved the problem with 2 trunked (Etherchannel in the cisco world) cat 6 cables between the units for heartbeat. (Originally the architects wanted only optical between racks to avoid it turning into spagetti junction)

 

 

Anyway, heatbeating over intermediate equiptment is asking for trouble and should be avoided at all costs!