Storage device redundancy with FlexFabric 5700

[Capricorn800]

Hi!

We are running two HP FlexFabric 5700 with IRF  as our core and then Aruba 25xx series for our mix distribution and access layer.

We are planning to buy 10G based storage and then connect them in a way to have redundancy. I did this on Cisco with VPC but need some information on HP and Aruba platform. 

One question is to connect the storage direct to the HP so that we have redundancy but I feel like its not the best thing to connect storage devices direct to core layer but let me know if its ok?

Can anyone point me to some reading material or suggestion how can I achieve redundancy like Cisco has this VPC.

Thanks

[bamaskas]

The 10G storage capabilities with regard to teaming the controller ports will drive how VPC, MLAG, or IRF are leveraged.  With two switches configured to look like one switch, storage controller connectivity must take advantage of this through connectivity.   Are you thinking of MSA, Nimble, 3PAR, or other storage system  ?

[parnassus · Accepted solution]

---

@Capricorn800 wrote:  One question is to connect the storage direct to the HP so that we have redundancy but I feel like its not the best thing to connect storage devices direct to core layer but let me know if its ok?

---

Why are you thinking that your Storage's NIC ports shouldn't be connected directly to your IRF? it will be way better than connecting them through distribution layer Aruba 25xx Switch series (if this is your alternative idea).

IRF provides resiliency and uplinking your Storage NIC 10G ports in a way (via LAG using LACP IEEE802.3ad) that originating Storage 10G ports are distributed equally on both IRF members is the best way to ensure resiliency and enhaced throughput.

IMHO a LAG of 2 (or 4) ports Storage side should be terminate into a LAG of 2 (or 4) corresponding ports on the IRF, with destination ports distributed on both IRF members (if 2 links are used: 1 port on IRF-1 and 1-port on IRF-2, if 4-links are used: 2 ports on IRF-1 and 2 ports on IRF-2)...that way your storage is protected by (one) IRF Member failure or by one (or two) ports failure (failures on a single IRF Members or distributed across the entire IRF).

The only case your storage (and not only it!) will be in trouble it will be when storage related 10G ports on IRF-1 will go KO and - concurrently - IRF-2 Member will go KO too (or viceversa)...but, at that point, you probably have other type of issues to solve if your core reports all those problems at the very same moment.

[Capricorn800]

Hi!

I thought of it as some old reading that talks more about Tier 3 and Tier2 with Core, Distribution and Access layer and Core connects only the switches down to Distribution so thought to double check and thanks for mentioning this direct.

The only issue I found is to not connect too many of them so that we will ran out of ports.

For future I am thinking to connecting  Aruba 3810 with HP 5700 as shown. Will this work?

[parnassus]

Yes, it will...BUT only if you're going to deal with just two 4 members LAGs per side (one 4 members LAG on IRF and the other 4 members LAG on VSF, inter-crossed together as per your schema)...it will not if you plan to deploy four 2 members LAGs as grouped on the drawing (that way they appear like four 2 ports LAGs).

[Capricorn800]

Thanks parnassus!

I am trying to undersand your point :)

[parnassus]

There is really nothing to understand: IRF and VSF are both virtualization technologies...it means that both will render (for the sake of simplicity) two physical switches to act and to be recognized as one...between two switches, generally, you don't want to create any loop...so between IRF and VSF there should be just one uplink made of some (I saw 4) aggregated ports (cross connected using the pattern drawn).

[Capricorn800]

I know about IRF VSF tech. They will act as one but the diagram I made was with Cisco VPC concept in mind.

[Capricorn800]

So you are refering to something like this?

[parnassus]

More to something similar to:

[parnassus]

---

@Capricorn800 wrote:

I know about IRF VSF tech. They will act as one but the diagram I made was with Cisco VPC concept in mind.

---

Cisco vPC is more similar - or comparable - to Aruba VSX (separate control planes), IRF is more similar - or comparable - to VSS. Aruba VSF (which use a frontplane stacking technology approach) is, on the other hand, more similar to IRF than to VSX...in any case two Aruba 3810M stacked are de facto using the backplane stacking approach which is similar to VSF...but implemented using specific dedicated Hardware (Stacking Modules and Stacking Cables)...that's to say that designing two Aruba 3810M in a stack is practically identical to design to Aruba 2930F/5400R zl2 in VSF (there are differences - in terms of different requirements and restrictions - but those are very stacking technology and switch series specific)...so in the diagram above you can use a (Aruba 2930F/5400R zl2) VSF instead of Aruba 3810M stack.

[Capricorn800]

Thanks. 

Can you recommend some location on HP site which such examples/configuration?

[parnassus]

Since HPE FlexFabric 5700 runs over Comware NOS while Aruba 3810M runs over ArubaOS-Switch (rebranded and enahanced HP ProVision NOS), for the sake of implementing LACP Trunks (LAGs) between a Comware (historically known as belonging to A-Series) and ProVision (historically known as belonging to E-Series)...follow this initial guide:

https://support.hpe.com/hpsc/doc/public/display?docId=emr_na-c03503743

Clearly you should adapt to IRF and Backplane Stack interface(s) numbering schemes since the example is referring to two standalone switches.

Edit: link above corrected.

[Capricorn800]

Thanks. 

Did you forward me the right link?

[parnassus]

Sorry, too many threads open concurrently and I cross-pasted! URL corrected.

[Capricorn800]

Thanks. Sorry for late replay as I was getting some error on login.

so the main thread was connecting storage direct to FlexFabric. So now I will create interface config for access port and will come back to you.