SOLVED
Magic_Hair wrote: I have 2 x Aruba 2930F which I want to make into a VSF Stack (via 10Gb ports)
That's good (so you should use 16.03.0003 software version, at least).
Magic_Hair wrote: I have a 3rd switch (Aruba 2540) which I want to trunk (i.e LACP or similar) to.Can I trunk using a single cable from each stack member to the 2540? (i.e two cables total) I just want to know if this is possible or if both cables have to come from the same stack member (former option has better redundancy obviously)
Since two Aruba 2930F deployed using VSF appear logically as one single logical switch to other switches (and hosts) you should only configure - on both ends - a Port Trunk (Link Aggregation Group) made with 2 (or more) physical links and configured to use LACP: that way (a) on the Aruba 2540 side the defined LAG clearly originates within the same single switch and (b) on VSF side the LAG you defined can be configured to use (it's a matter of resiliency againsta a VSF Member failure) physical links located on each/different VSF Member (e.g. 1/0/x and 2/0/x ports where x is the port on each VSF Member and 1 and 2 are the VSF Members IDs)...in both sides the LAG you are going to define is a normal Port Trunk=LAG=BAGG (it's not Distributed-Trunking).
Clearly you should evaluate IF that topology is useful...considering also how/where the hosts (and servers) are distributed on VSF side which should act as the "core" of your network (e.g. the same approach should be applied to each Server <--> VSF Stack inter-connection, for the same reasons).
I'm not an HPE Employee
Magic_Hair wrote: I partially get it. You are saying I should be able to create a trunk to a 3rd switch (2540) and cable from both VSF members.
Exactly. You should create two Link Aggregation Groups or LAGs (also called Bridge AGGregations or BAGGs) which are the entities through which you define your Port Trunks, a LAG per Switch...in this case a LAG on the 2540 and a LAG on the VSF Stack (which acts as one logical Switch made of two physical Switches). There isn't any Distributed-Trunking going on (so no Inter Switch Link ISL) because the VSF Stack is already (and acts as) one single logical switch.
Magic_Hair wrote: This guide link pg120 says there are two main types of trunks: "LACP" & "Trunk". Which one did you suggest I use? (and I presume vlans can go down either?
There is some "confusion" about the Trunk term: it is used to define a Port's Link-Type and also to define a LAG type (so in "Port Trunking").
A port (also a logical one like a Trunk=LAG one) can be set with Link-Type equal to "Access", "Trunk" or "Hybrid" (look for HPE KB about that).
Conversely, when you speak about Port Trunking (so LAG=BAGG which describe the same concept) you have basically two main choices: Port Tunking deployed using no protocol (here the culprit: called - again - Trunk) or Port Trunking deployed using Dynamic LACP (IEEE 802.3ad), which is very often the preferred configuration option when you deal with LAGs.
Magic_Hair wrote: This guide link pg23 says that a VSF stack cannot use Distributed Trunking - when I look up what this is, (definition) it sounds exactly like what I am trying to achieve - hence my confusionWhat is the definiation of BAGG? I am not familiar with that acronym
That guide is right: VSF doesn't support Distributed Trunking...but if you read what I wrote above you discover that you don't need DT. VSF lets you to use a normal trunking between the VSF Stack and your 2540 Switch and you're free to use, VSF side, two (or more) LAG member ports residing on the same VSF Member (any of the two, three or four ones) or to be spread across VSF Members (like I suggested).
Re: Servers - yes they will have multiple NICs and they will cable across members
OK, that's another interesting application: a Server NIC which has 2 or more ports can be used to team those ports (again, with Dynamic LACP) and then link those ports against the VSF Stack in the same way you do with the 2540.
For bonus points, I will need to setup a MAD device (to control VSF split brain) I want this to be the 2540 - I haven't studied this yet, but will want to know that the accepted design allows this.
VSF requires a Multi-Active Detection (MAD) device in order to manage/mitigate VSF Split-Brain events...but device used should support LACP-MAD and LLDP-MAD too [*], not only LACP and LLDP (at least, from what I understood), and that requirement is essential because you do special Port Trunking between the MAD and the VSF Members (side note: it seems that LLDP-MAD works up to 2 VSF Members...).
It's a pity that Aruba 2930F doesn't support an OoBM (Out of Band Management) port since exactly OoBM ports can be used to deploy MAD mechanism (Aruba 2920 has a front-facing OoBM port).
I've read (search Aruba AirHeads Community, information is a little bit fragmented) that an Aruba 5400R zl2 based VSF Stack can be deployed with a MAD mechanism built up through VSF Members OoBM ports (which is nice because it doesn't require a Switch supporting specifically LACP-MAD to act as a MAD Device).
[*] really the MAD device requirements look a little bit less restrictive (no reference about LACP-MAD [**]):
- A MAD assist device must have support for LACP (IEEE 802.1AX) LAG interfaces.
- It should be SNMPv2 enabled and community information must be configured on the VSF device as part of MAD configuration.
- It should have support for LLDP (IEEE 802.1ab rev) and the basic management TLV set as defined there in.
- It should support SNMP GET access to the LLDP remote MIB (IEEE 802.1AB D13) and the ifTable MIB (RFC 2683), Aruba switches have LLDP enabled by default.
- Support for ARP is required.
[**] LACP-MAD is implemented by sending extended LACP Data Units (LACPDUs) with a Type Length Value (TLV) that conveys the active ID of an VSF virtual device.
I'm not an HPE Employee