SOLVED
Hello @lasse-k are you able to show us sanitized running configuration of the Aruba 5412R zl2 (sanitized = obfuscated MAC addresses, Serial Numbers and other sensitive details of your devices/networks)?
To complete you would provide the outputs of show modules, show tech transceviers, show interface ethernet x (with x=I1, J1, I2 and J2) and show vlan port trk101 detail (and trk102, eventually) too, it would help a lot to exactly understand what is going on.
Also, there is something strange on DK1-SWITCH01: you reported that port I2 and port J2 are members of Trk101 (the command reported is indeed trunk I2,J2 trk101 lacp) but, on the next show lacp command's output, the Trk101 logical interface looks like was made of port I1 and port J1 instead of I2 and J2. Isn't it?
Side note: is the Aruba 5412R zl2 running an updated version of its ArubaOS-Switch?
Would be nice to also have a clear picture of what are materials (SKUs) actually involved...are you using QSFP+ Transceivers or DACs? Are you able to report us back the SKUs involved on both sides (HPE 5710 and Aruba 5412R zl2)?
I'm not an HPE Employee
The fact that QSFP+ MPO SR Transceiver is not recognized on Port I1 (Aruba 5412R zl2: Module I Port 1) is not a good sign.
Your Aruba 5400R zl2 Switch should be running using the ArubaOS-Switch KB.16.05.0007 (Mid 2018)...not new but also not too old. Definitely not ancient. Using a newer version would probably be better but, technically speaking, used modules and QSFP+ MPO Transceviers are supported.
Is your Aruba 5400R zl2 operating in v3 only mode or is operating in Compatibility mode?
You have to know the Aruba 5400R zl2 is able to operate in two modes:
- Compatibility mode: This is the default mode and it allows the Aruba 5400R zl2 chassis to run both v2 zl and v3 zl2 modules concurrently, but the software feature set will be equivalent to v2 zl (as example it cannot enable any v3 zl2 specific software feature).
- V3 only mode: Some features are supported in v3 zl2 only mode so you need to “turn off” the v2 zl modules. When moving from the default v2-compatible mode to v3-only mode, the switch reboot is required.
Check this. Not sure about the impact of runnning in Compatibility mode (v2 zl/v3 zl2) since the J9996A is for sure v3 zl2 but, AFAIK, it's recognized on your chassis so you should be OK. Nevermind, do a check.
The only QSFP+ Transceiver I see recognized by your Aruba 5400R zl2 is the one inserted into Module J Port 1 and it is exactly a JH231A with 4x4 1999-4554 (That's the SKU).
As of now the Aruba 5400R zl2 supports those QSFP+/DAC once installed on QSFP+ ports of modules with SKU J9996A (it's your case) and J9992A (it's not your case):
- HPE X142 40G QSFP+ MPO SR4 Transceiver (JH231A)
- HPE X142 40G QSFP+ LC LR4 SM Transceiver (JH232A)
- HPE X142 40G QSFP+ MPO eSR4 300M Transceiver (JH233A)
- Aruba 40G QSFP+ LC Bidirectional 150m MMF 2-strand Transceiver (JL308A)
- HPE X242 40G QSFP+ to QSFP+ 1m Direct Attach Copper Cable (JH234A)
- HPE X242 40G QSFP+ to QSFP+ 3m Direct Attach Copper Cable (JH235A)
- HPE X242 40G QSFP+ to QSFP+ 5m Direct Attach Copper Cable (JH236A)
So you're dealing with an MPO SR4 QSFP+ Transceiver on the Aruba side.
On the HPE FlexFabric 5710 you're correctly using the corresponding JG325B MPO SR4 Transceiver, which is supported on the HPE 5710:
- HPE X140 40G QSFP+ MPO SR4 Transceiver (JG325B)
- HPE X140 40G QSFP+ MPO MM 850nm CSR4 300m Transceiver (JG709A)
- HPE X140 40G QSFP+ LC BiDi 100m MM Transceiver (JL251A)
- HPE X140 40G QSFP+ LC LR4 SM 10km 1310nm Transceiver (JG661A)
- HPE X140 40G QSFP+ LC LR4L 2km SM Transceiver (JL286A)
- HPE QSFP/SFP+ Adapter Kit (655874-B21)
- HPE FlexNetwork X240 40G QSFP+ QSFP+ 1m Direct Attach Copper Cable (JG326A)
- HPE FlexNetwork X240 40G QSFP+ QSFP+ 3m Direct Attach Copper Cable (JG327A)
- HPE FlexNetwork X240 40G QSFP+ QSFP+ 5m Direct Attach Copper Cable (JG328A)
- HPE FlexNetwork X240 40G QSFP+ to 4x10G SFP+ 1m Direct Attach Copper Splitter Cable (JG329A)
- HPE FlexNetwork X240 40G QSFP+ to 4x10G SFP+ 3m Direct Attach Copper Splitter Cable (JG330A)
- HPE FlexNetwork X240 40G QSFP+ to 4x10G SFP+ 5m Direct Attach Copper Splitter Cable (JG331A)
- HPE X2A0 40G QSFP+ to QSFP+ 7m Active Optical Cable (JL287A)
- HPE X2A0 40G QSFP+ to QSFP+ 10m Active Optical Cable (JL288A)
- HPE X2A0 40G QSFP+ to QSFP+ 20m Active Optical Cable (JL289A)
Couldn't be an issue related to the MPO cabling used? are you going directly (FO patch) between two Switches?
If you’re going QSFP+ MPO to QSFP+ MPO directly, are you using an MPO Female to MPO Female crossover fiber optic patch cable or what?
Is it a Key-Up to Key-Up crossover MPO 12-strand cable?
Other question: could you provide (R|M)STP - Spanning Tree - status on both ends (Aruba/HPE)?
Also, can you post details about Comware V7 version currently running on your HPE FlexNetwork 5710 (SKU: JL586A) Switch? <- use the display version CLI command.
Are you sure that, HPE 5710 side, the six QSFP+ ports are all in their default state (not splitted 40Gbps -> 4x10Gbps)?
I noticed there is this particular note about JG325B but it looks not too relevant.
Edit: ironically the latest (6th May 2020) HPE Comware 40G Transceriver doesn't list the HPE FlexNetwork 5710 (JL586A) Switch among those supporting the QSFP+ Transceiver JG325B...the (4th May 2020) HPE FlexNewtork 5710 QuickSpecs instead correctly reports the JG325B as normally supported.
I'm not an HPE Employee
What I'am trying to set are two rack switches for servers using 2 x HPE 5710 switches. The switches will both connect to the main switch Aruba 5400. All server NIC's will have connections to both HPE 5710 switches.
Both HPE 5710 switches have 2 cables connected to the Aruba 5400 switch in a trunk.
Somehow I got the trunk up running no on the first switch and will now try configure the other one.
I'am a bit in doubt about the spanning-tree setup on these switches so any suggestion will be appreciated a lot. Might be the whole issue / problem that the spanning-tree setup are done wrong?
Hi! Glad you fixed (just reseating QSFP+ Transcerivers on the Aruba 5400R zl2?) the LAGs that are now working on both peers (at least from your Aruba 5400R zl2 to one of your HPE FlexNetwork 5710).
Just a question: why not deploying IRF with your two HPE FlexNetwork 5710? that way you would use Port Trunking (LACP) on Aruba 5400R zl2 with member ports going from the Aruba chassis to both 5710 chassis (so you will "distribute" the physical links to both your 5710 concurrently)...one single Trk<x> on the Aruba side and one single Bridge Aggregation interface (LAG) on the HPE 5710 IRF side.
The same could be done (should be done, at that point) between the Server and the HPE 5710 IRF pair.
Spanning Tree should be configured accordingly (in any case)...so you should decide what Spanning Tree protocol you want (RSTP/MSTP) and then set both peers (here I'm supposing you would go with the IRF approach...so the IRF become a single logical peer for the Aruba 5400R zl2 despite is made of two physical switches).
I'm not an HPE Employee
@lasse-k wrote: Lets consider I will do a "normal" trunk setup with the 2 switches as shown in the small diagram. Can I them just like the IRF setup make 1 trunk on the Aruba switch for all connections from both switches?
If your HPE 5710's pair will remain made of separated (not uplinked together) switches it will just a matter of setting up two separated trunks (as per your network diagram) with links egressing from the Aruba 5400R zl2 and ingressing, respectively, into each one HPE 5710.
Say:
Aruba 5400R zl2 trk<x> (2 links) <===> (2 links) HPE 5710 (East)
Aruba 5400R zl2 trk<y> (2 links) <===> (2 links) HPE 5710 (West)
Instead of (IRF case):
Aruba 5400R zl2 trk<z> (e.g. 4 links) <===> IRF (say 2 links on 5710 East and the others 2 links on the 5710 West)
With regards to Spanning Tree: yes, even if the network topology hasn't any loop (your one, as per diagram, hasn't <- provided that the Server is not going to bridge its ports), it's best practice to enable and configure Spanning Tree accordingly because STP kicks in to protect your network from loops.
I'm not an HPE Employee
@parnassus wrote:With regards to Spanning Tree: yes, even if the network topology hasn't any loop (your one, as per diagram, hasn't <- provided that the Server is not going to bridge its ports), it's best practice to enable and configure Spanning Tree accordingly because STP kicks in to protect your network from loops.
Just to be sure. My server will indeed have 2 bridged ports and supposed to have 1 cable in each switch. If this setup have to run I will need the Spanning tree setup - right? Putting the Aruba 5400 as the root bridge and the others with lower priority. And will the Aruba be ok to "speak" with the HPE devices?
There shouldn't be issues between Aruba and HPE in terms of STP. Let me check if I'm able to find some documents about STP interoperability between ArubaOS-Switch (ProVision) based devices and Comware based devices.
With regards to the Server: I mean...if the two ports are set as switched on the very same VLAN the Server is going to form (and close) a loop on that VLAN so I expect STP to kick in and disable one offending link (between the Server and the IRF). Isn't it?
I'm not an HPE Employee
If the NIC ports are "bonded" together (via "LACP" = Dynamic Trunking or via "Non Protocol" = Static Trunking) [*] then STP will see those ports as one entity, the logical interface represeting the Link Aggregation Group (LAG) so, correctly, it will not kick in since the Switch will see one aggregated port and not two physical interfaces (from the STP standpoint).
But you wrote that NIC ports will be "bridged" and I associate the bridging action with something which is not the aggregation we are here referring to so my note above about the possible loop.
[*] in your scenario, not having an IRF deployed, you can't "bond" NIC's ports together and terminate their links against different standalone 5710...so, at best, in this conditions, each NIC port can only be configured to be simply "standalone" (without any traffic bridging between the them).
I'm not an HPE Employee
Hello
Hmm - ok. Seems I need to go the IRF way then.
Tried to make a diagram below. When setting up a IRF I assume the one switch can break down and I will still be up running? Do one of the switches act as primary and if this breaks down the second switch will also be down?
I will then connect the switches like below - will this be possible? To ensure redundancy.
The issue about the NIC / LACP / static trunking will then be gone I assume as both switches will then act as one?
