Aruba & ProVision-based
cancel
Showing results for 
Search instead for 
Did you mean: 

LAG/LACP trunks and spanning tree on 2910al-24g

 
SOLVED
Go to solution
pauladam2608
Occasional Advisor

LAG/LACP trunks and spanning tree on 2910al-24g

Hi All,

Firstly I will say for the record, I come from a Cisco IOS background :)

Ive been landed in a situation where I have a set of 2 HP 2910al-24g switches with a 4 port Trunk (LAG/LACP) between them and I need to add a new VLAN to the trunk group....simple enough, however, not being that familiar with HP switches I need to get a handle on the ramifications of making changes to the trunk port and the spanning-tree settings that are associated.

The easiest thing to do is show the relevant parts of the config as a start....

trunk 21-24 trk1 trunk

interface 21
flow-control
exit
interface 22
flow-control
exit
interface 23
flow-control
exit
interface 24
flow-control
exit

spanning-tree
spanning-tree Trk1 priority 4
spanning-tree config-name "MyConfig"
spanning-tree config-revision 1
spanning-tree instance 1 vlan 1 10 20
spanning-tree instance 1 priority 0
spanning-tree instance 1 Trk1 priority 4
spanning-tree instance 2 vlan 30 40
spanning-tree instance 2 priority 1
spanning-tree instance 2 Trk1 priority 4

Thats everything that is relevant as far as I can see - the switches are not stacked, they are individual units, but both have the same config.

I need to add VLAN 64 to the Trk1 trunk - which of course is straightforward to do in terms of adding the "tagged Trk1" config to the vlan - but what I dont know about is adding it into the spanning-tree instances - particularly whether adding it will cause the trunk to drop and renegotiate etc.

I dont understand why there are two configured spanning tree instances either? (I didnt do the config, like I say, I have just inherited it) - is it a case that I can just add VLAN 64 to one of the instances without interruption, or should I create a new instance for my new VLAN without causing a drop? Or am I facing a drop no matter what I do? 

And before anyone says - Ive renamed the spanning-tree config name in the snippet above...but does that have any bearing in sharing of the config, or is it just an identifier, I assume all switches must have the same config name?

Any help/advice greatly appreciated.

Thanks

Paul

16 REPLIES 16
parnassus
Honored Contributor

Re: LAG/LACP trunks and spanning tree on 2910al-24g


pauladam2608 wrote: I dont understand why there are two configured spanning tree instances either?

Good question to starting with.

The point is it looks like it was configured exactly to prevent Spanning Tree to activate if those Switches (supposing they were the only ones involved in your scenario) have (or have had) more than one active link between them (so as they hadn't a Port Trunk configured since beginning...so active links formed a physical loop managed by MSTP with its spanning tree instances).

But then the question is: if actually really there is just a 4 ports LACP Port Trunk what's the reason for those MSTP instances?

What's about providing show lacp, show trunk and show vlan outputs? ...just to be sure about Trk1 and VLANs.

As you said adding the VLAN Id 64 to Trk1 Port Trunk is pretty straightforward...as doing:

vlan 64 tagged trk1

 

pauladam2608
Occasional Advisor

Re: LAG/LACP trunks and spanning tree on 2910al-24g

Here's the outputs (and a some more of the background below also)

sh lacp

no LACP ports found (???)

sh trunk

Load Balancing

Port | Name Type | Group Type
---- + -------------------------------- --------- + ----- --------
21 | 100/1000T | Trk1 Trunk
22 | 100/1000T | Trk1 Trunk
23 | 100/1000T | Trk1 Trunk
24 | 100/1000T | Trk1 Trunk

sh vlan 

Status and Counters - VLAN Information

Maximum VLANs to support : 256
Primary VLAN : DEFAULT_VLAN
Management VLAN :

VLAN ID Name | Status Voice Jumbo
------- -------------------------------- + ---------- ----- -----
1 DEFAULT_VLAN | Port-based No No
10 vMotion | Port-based No No
20 iSCSI | Port-based No Yes
30 vSphere | Port-based No No
40 FaultTolerance | Port-based No No
64 HQServersNew | Port-based No No

So as you can see this is a pair (ok...so its actually two pairs! - I'll explain further below) of switches handling a lot more than it probably should be (VMware vMotion, iSCSI, FT and VM networking) - but it is what it is.... the VMware hosts are split across the two switch chassis, hence why the trunk is in place between the two - note this is not the way I would normally do this in my traditional Cisco background, it would be a stacked pair with Etherchannels for the VM networking for a start with IP hashing on the VMware vSwitches, but I dont have the luxury of being able to change all of this on a whim Im afraid - also, currently the entire LAN is residing in VLAN 1 (yes...I know - again, its inherited!) and is running on a 10.0.0.0/8 (facepalm!), so I am working towards segregating things up into seperate VLANs and stage 1 is to get the servers into their own VLAN - as you can see...VLAN 64
But hopefully, based on this, you can see why I am nervous about what may happen if I change any spanning-tree parameters on that trunk link (I was surprised to see no LACP to be honest - but there you go, thats the lack of my knowledge of the HP switch OS).

As I said its actually two pairs (probably a KEY bit of info I should have mentioned!!) - the switches each have the HP J9008A 2-port 10-GbE SFP+ module installed - the VMware environment is split into two locations within the premises.

So switches 1 and 2 are in the same location and connected with the 4 port trunk as previously mentioned and switches 3 and 4 are exactly the same, however switches 1 and 3 are connected by a 10g fibre and switches 2 and 4 are also connected by 10g fibre - hence...a BIG loop - so this is the critical bit of info regarding the spanning tree, sorry about missing that - was trying not to do an information overload in the first post :) - I've added a quick screen snip of a diagram to illustrate the connectivity

So hopefully you now see why I need to add the new VLAN into the spanning-tree config - but it comes back to the original question of will I suffer a drop in connectivity? (with iSCSI and critical VM's communicating over this, I need to be sure!)

Thanks again

Paul

parnassus
Honored Contributor

Re: LAG/LACP trunks and spanning tree on 2910al-24g

Hi Paul, now the scenario is a more complete.

No LACP because the Port Trunking(s) you have are all using the Non-Protocol Port Trunking mode instead of the LACP (IEEE 802.3ad)...things were done that way but who know what was the reason (I suspect that was used "Trunk" - No Protocol - as Port Trunking type instead of Dynamic - LACP - due to MSTP requirement but I could be totally wrong on that). Doesn't matter for what you want to achieve.

I think you need to learn - on all four involved Switches (SW1, SW2, SW3 and SW4) - what is the MSTP configuration (Instances) to understand how those Instances were configured (it looks like you have Instance 1 with VLAN IDs 1 Default, 10 vMotion and 20 iSCSI and Instance 2 with VLAN IDs 30 vSphere and 40 Fault Tolerance)...so traffic tagged on those VLAN IDs will obey to MSTP configuration and use different links to reach involved hosts eliminating, from the point of view of each instance, the physical loop we see (my take: Instance 1 sees a path loop-free for VLAN Ids 1, 10 and 20...the same happens for Instance 2 for VLAN Ids 30 and 40).

Probably it's only a matter of adding the new VLAN Id 64 to the proper instance giving that you know what instance that VLAN should be part of and the path it will use (depending on being part of a particular MSTP Instance other than another).

As reference see the HP Switch Software Advanced Traffic Management Guide W.15.14 (from Page 70, MSTP) and, if necessary to basic management of the Switch, the HP Switch Software Management and Configuration Guide for W.15.14. On the first guide, at Page 112, MSTP is finally described so if I were you I will start from there (that guide starts immediately with MSTP configuration which is IMHO a little bit discouraging without knowing what's the MSTP main purpose), if you want to avoid jumping on that page give a read to this other chapter (it's about another Switch family but the subject is always about MSTP and its description is correctly provided at very beginning).

pauladam2608
Occasional Advisor

Re: LAG/LACP trunks and spanning tree on 2910al-24g

Hi Parnassus,

Thanks for getting back to me - yes, apologies for being "light" on the detail in the first post - I was trying to summarize, but when you mentioned about the loop I realised I had to provide the full detail :)

Anyway, I do understand the detail surrounding spanning-tree and whats its there for etc - been involved wth networks for a very long time, just not HP (sorry :))

I just dont get why the multiple instances for this - as far as I can see they are defining exactly the same loop for both (I will go anbd check though), which brings me right back to my original question.....(sorry - I will go and read the docs you sent, but I did do some reading before posting and couldnt find a definitive answer).

Anyway, this is still the main question - Will adding my vlan 64 to either of those instances cause a full reconvergence to occur - i.e. a drop in connectivity? Could I create a new instance for just VLAN 64 and not effect instances 1 and 2?

Thanks again

Paul

parnassus
Honored Contributor

Re: LAG/LACP trunks and spanning tree on 2910al-24g

Mmm...I don't believe that adding a new VLAN Id (64) on an existing MSTP Instance will cause any convergence (re)calculation (should verify but topology doesn't seem to change by adding that VLAN)...so that operation should not - di per sè - drop anything (clearly it all boils down to what ports that VLAN 64 it will be tagged to).

If those two Instances share the same path (and both block the same link) what's the purpose of having two instead of just one? having two identical will mean that a 10G (up)link between those two pairs will never be used (would be interesting to see Path(s) costs to understand better the whole configuration), isn't it? ...supposing the unlocked path shared by both instances uses one of those two 10G uplinks or the other but, clearly, not both.

TerjeAFK
Respected Contributor

Re: LAG/LACP trunks and spanning tree on 2910al-24g

I'm only guessing of course, but could the two MSTP instances have been used as load balancing, ie. that instance1 would use two of the uplinks while instance2 used the other two?

parnassus
Honored Contributor

Re: LAG/LACP trunks and spanning tree on 2910al-24g

Hi @TerjeAFK, nice to have another pair of eyes looking at this.

But...if so...why the priority is the same (4) WRT Trk1 (that particulat Trk1 <-- what Switch is it referring I miss, also we miss all the others on remaining Switches, isn't it?):

spanning-tree instance 1 Trk1 priority 4

spanning-tree instance 2 Trk1 priority 4

I mean the whole Trk1 has the same priority in both MSTP Instances...so I doubt Trk1 member ports can act differently (in a sort of load splitting)...or is that possible?

I expected to see something related to those two 10G uplinks...let me say: Instance 1 use one path blocking the other, Instance 2 the vice-versa...so, as example, we should see two free loop paths:

  • SW2<=>SW1<--10G Uplink 1-->SW3<=>SW4 when SW2<--10G Uplink 2-->SW4 is blocked on a MSTP Instance
  • SW1<=>SW2<--10G Uplink 2-->SW4<==>SW3 when SW1<--10G Uplink 1-->SW3 is blocked on the other

Or, guessing, both 10G uplinks are always permitted in any instance and, instead, 4x1G Trunk(s) - with lower path cost - between SW1/SW2 and SW3/SW4 alternate WRT MSTP instances, so we should see something like:

  • SW1<--10G Uplink 1-->SW3<=>SW4<--10G Uplink 2-->SW2 when SW1/SW2 Trunk is blocked on a MSTP Instance
  • SW3<--10G Uplink 1-->SW1<=>SW2<--10G Uplink 2-->SW4 when SW3/SW3 Trunk is blocked on the other
TerjeAFK
Respected Contributor

Re: LAG/LACP trunks and spanning tree on 2910al-24g

I agree if the whole STP config has been posted. So to come back to the original question: pick one of the instances (best guess) to add the new vlan as tagged. But you should look into the config and if possible change it to using dynamic trunk (LACP) and a single MSTP instance (or switch to RSTP) if there are no special needs.

 

pauladam2608
Occasional Advisor

Re: LAG/LACP trunks and spanning tree on 2910al-24g

Sorry for delayed response, been dealing with some other things today so far....

Im going to pull the full config from all 4 switches and some additional "show" commands for you to have a look at, bear with me - I'll post shortly

I agree with looking at LACP, but I just dont have that luxury just now (its a long story, but Im only here temporarily and trying to do the best I can with a non standard situation (to my eyes anyway)