5900 Switchs IRF and 6125XLG Switchs IRF with long distance

[fabbb]

Hi all,
I have a question on HPE Switch cluster (IRF) with Switchs located on different physical sites.
I didn't find the anwser in HPE docs.

As the sites are pretty far from each other, I plan to use 10G ER SFP+, which support 40km long fibers. Fibers will be about 20km long, between 2 switches of the IRF..
Do you think clustering will work with such a distance ?

There will be 2 IRFs with that config : one IRF with 5900 Switchs and one IRF with 6125XLG Switchs.

Thanks !

[parnassus]

The description isn't very clear: are you asking if a single IRF Stack (made of multiple similar IRF Member switches, all 5900 OR all 6125XLG, not a mix of them) can be formed interconnecting one IRF Members pair on Site A with the other IRF Member pair on Site B where the interconnection happens through straight (end-to-end without any inbetween switch) 10G Fiber Optic links? Yes, it can.
But you can not form an IRF Stack made of a mix of 5900 and 6125XLG...you can't, as example, create a single IRF Stack where two 5900 are on Site A and the others two 6125XLG are on Site B: That's not possible (IRF Stack requires - generally - similar IRF Members switch series to be formed)
If, instead, your scenario is: two separated IRF Stacks (an IRF Stack 1 made of two 5900 on Site A and, geographically separated "n" km/miles away each others, another IRF Stack 2 made of two 6125XLG on Site B) then you're not literally stacking them together...you're only just interconnecting them together (uplinking them with one or more F.O. physical links)...and, again, you can do that...provided that you create - I'm presuming you want to do that given the number of F.O. links you have available - a LAG on each end (so you can aggregate both F.O. physical links on a logical aggregated trunk) and use that LAG to let IRF Stack 1 and IRF Stack 2 to be linked OR provided that you're using a single link uplink... in any case distance issue is won using right LR/ER transceivers.
Clearly uplinking is different than setting up a single IRF Stack with some geographically separated IRF Members (as a possible scenario described above).

[fabbb]

Hi Parnassus, and thanks for your feedback!

I have currently 2 (working) clusters per site : on each site : one cluster formed with 2x 5900, and one cluster formed with 4x 6125XLG. On each site the 2 clusters are connected together for traffic.  The 2 sites are communicating with classical routing via routers/firewalls.

The plan is to merge the 5900 clusters, and to merge the 6125XLG clusters.

As drawing is always better than talking, target config is shown in diagram attached (ToR-SW=5900 , BL-SW=6125XLG) :

The long distance fibers I plan to add are the 8 between the sites (site A = CIT, site B = PK5).

As per as your answer I believe the new clusters will be OK , with ER transceivers on the new fibers.

I told customer (who is going to install the new inter-site fibers) that the L1 (physical) continuity is required : only passive 'equipments' in the middle , so that if one port is unplugged on one switch in site A, the other end on site B goes down. Or if multiplexing (WDM., etc) is used between sites, it must configued so that L1 (physical) continuity is achived.

I did not find such clear requirement (physical continuity between switches on different sites) on HPE doc, but I believe this is required for proper IRF building (in particular redundancy inside IRF ports), isn't it ?

Thx,