2920 LACP Redundancy Question

[Mirthrock311]

I have two stacks of HP Aruba 2920-48G-PoE+ switches.  One stack is in our MDF and the other stack is in the IDF about 200 feet away.  The first switch in each stack has an SFP+ module installed with a fiber transceiver.  The two stacks are connected via an LC/LC fiber cable.  We also have a four Ethernet runs between the MDF and IDF which are not being used.  I'd like to create a Trunk Group to utilize the four Ethernet runs and the Fiber run.  I was able to add the fiber and four Ethernet runs into a LAG using LACP.  I was curious if using dissimalar media is supported.  Will I run into any issues by configuring the switches to use Fiber and Ethernet in the same LAG?

Thanks!

[parnassus]

Using dissimilar media type is not supported in the same port trunk group (see, as reference, here or in the Aruba 2920 Software Management and Configuration Guide):

"The switch does not support port trunking through an intermediate, non-trunking device such as a hub, or using more than one media type in a port trunk group. Similarly, for proper trunk operation, all links in the same trunk group must have the same speed, duplex, and flow control."

LACP requires full-duplex (FDx) links of the same media type (10/100Base-T, 100FX, 1GB and so on) and the same speed, it enforces speed and duplex conformance across a trunk group...and this is valid also if you're going to configure Backup Links within the LAG which is using LACP Dynamic (I mean: up to 8 operating ports can be allowed per trunk but one or more additional port can be defined as backup ports that LACP will automatically enable if one primary port fails).

This doesn't mean you can't create two separate LAG(s), one with fiber links (supposing you can add another SFP+ Transceiver to your actual one and also supposing you have another fiber optic run to use) and the other one with Ethernet copper links, but then you can't use both LAG(s) to concurrently link your two Aruba 2920 stacks (at least I think this sort of double inter-switch link is going to be useless, as both trunks will form a loop between the two stacks and one will be disabled by STP/RSTP).

[TerjeAFK]

If you use MSTP you could potentially load balance VLANs between the two LAGs, but this is tricky to setup and maintain and is not for the faint-hearted :-)

[parnassus]

Supposing your Aruba 2920 Switches are yet equipped with, at least, two HPE Aruba 2920 2port 10GbE SFP+ (J9731A) modules and considering both Aruba 2920 Stacks (Ring Topology, I hope) presence and the distance between them probably one simple and efficent addition you can think of (expensive considering necessary additionals J9731A Modules and SFP+ Transceivers plus a new Fiber Optic line) is to:

• add a SFP+ Module and SFP+ Transceiver per Stack on the actual 2920 which doesn't have it yet to have a 2nd 10G port.
• use a new Fiber Optic run
• aggregate on a LAG two 10G ports coming from different 2920 within the same Stack (on each location)
• use LACP Dynamci on that LAG.
• use that LAG (made of one 10G from 1st 2920 + one 10G from 2nd 2920) to connect to the other Stack (with identical configuration) on the other location.

Doing things that way you have redundancy:

• at Module level (since you don't use the same J9731A Module on the same Aruba 2920 Switch within the same Aruba 2920 Stack)
• at Fiber Optic links level (since your LAGs will survive to a single Fiber Optic link failure)
• at SFP+ slot level (since if a SFP+ slot of the J9731A Module fails you can re-arrange your Hardware/Software configuration to use the remaining unused SFP+ slot on the same J9731A Module).
• at Backplane Slot level (since if you have issues on those Slots on one Aruba 2920 you can always move the J9731A Module on another one Aruba 2920 within the same Stack).

I know...it's quite expensive, it is more expensive than using the actually installed single J9731A Module and enable its second SFP+ Slot port to create the 10G+10G LAG discussed above.

In both cases you will have better performances between Stacks than using just a single 10G uplink (it's not a Trunk, it's just a single uplink with the port set Trunk Type, if necessary) or 4x1G LAG.

[Mirthrock311]

Thank you all very much for the replies.  This is very good information.

 I think ultimately I will get another SFP+ module and run another fiber line so that I can configure the fiber lines to be their own Trunk.  But in the meantime, I'm considering using the current 10Gbps Fiber run as the primary connection between the two stacks and then configure the 4x1Gbps in a trunk to use a backup connection.  I'm assuming there is a way to set the weight of the two paths so that if the fiber line goes down, the 4x1Gbps trunk will start passing traffic?

Again, thank you all for the help.

[parnassus]

Not an expert speaking here...and also not sure if this idea is actually superseded...but probably the type of uplink redundancy that you're looking for (and that you're somewhat forced to deploy) can be managed using something called Fast-Uplink STP or Fast Uplink Span (a concept that was adopted on STP compliant ProCurve switches many years ago and that was related specifically to STP, not RSTP)..probably it's not the same feature that is now called STP Fast Mode (supported on Aruba 2920) which could have been called Fast Port Span.

Probably with the Fast Uplink STP/Fast Uplink Span technique you will be able to set the 10G port to forward traffic (acting as the STP root upstream port) to a STP root Switch (the other VSF stack that should run STP too...) while the traffic on the four 1G ports Trunk Group will remain blocked down by STP...if the 10G link will fail the system (your VSF in this case) should be able to do the STP reconvergence and enable the traffic forwarding on the previously excluded Trunk Group...this in few seconds (about less than 10"?).

Would be interesting to know if working on STP Port Priority (lower value decides who is forwarding) and STP Port Mode (Uplink mode on involved ports/trunk) of the upstream ports (the 10G and the Trunk of 4x1G) will suffice or if with the use of VSF there is another better way to achive the type of uplink redundancy you want with the interfaces and physical links you have now.

P.S.

I'm not sure if the Aruba 2920 supports the Fast-Uplink STP (or the Fast-Uplink MSTP)...so all I wrote above could be quite useless. Hope it's not.

[parnassus]

Or...trying the Aruba 2920's SmartLink feature:

"Smart link is a switch feature that provides effective, simple and fast-converging link redundancy in network topology with dual uplink between different layers of the network. It requires an active (master) and a backup (slave) link. The active link carries the uplink traffic. Upon failure of the active link, a switchover is triggered and the traffic is directed to the backup link."

even if I'm not sure (I've no direct experience about that) if SmartLink can be correctly depolyed between two Aruba 2920's Stacks because I'm in doubt that your two Stacks can be seen as "different layers of the network".

See the "HPE ArubaOS-Switch Advanced Traffic Management Guide for WB.16.01" and use at least ArubaOS-Switch version WB.15.18.0010 or newer (the WB.15.18.0011) because with software releases before the WB.15.18.0010 there is an issue between SmartLink and Stacking (the "CR_0000190943 Symptom: Traffic is not properly forwarded through smart link ports on switches configured in stacking mode.")...at least it means that SmartLink can be deployed on a Stack...the point is to understand if it can also work between two Stacks and if it admits Bridge Aggregation groups as links.

[Mirthrock311]

Thank you all for your help with this.  I've decided to investigate HP's Smart-Link technology as a method of failover.  You can see my trials and tribulations with Smart-Link in this thread.  If I cannot get Smart-Link to work properly, I will investigate Fast Uplink STP.  I will update this thread again once I have a solution.  Thank you!