VLANs Across Switches With LACP Trunks

[wtu]

Hi There,

I'm new to the HP Procurve world. Could anyone help me on configuring Procurve 2848 switches with VLANs across multiple switches and LACP trunks presence? I couldn't find any examples from manuals. Below is the situation:

There are three 2848 switches (A,B,C) with three VLANS (VLAN 1, VLAN 100, VLAN 200). A has four ports configured as LACP Trk1 connecting to B (as Dyn1 showing on B). B has two ports configured as LACP Trk2 connecting to C (as Dyn3 showing on C). The IP routing is enabled on A but not on B or C. Trunking ports are tagged.

Before I configured LACP trunking, routing and VLANs seemed to work fine since pinging from workstation on one VLAN to another workstation on another VLAN worked.

However, after I configured trunking, the pinging connectivity broke. Is there any examples? Thanks!

[Jody_Greene]

I would recommend not using LACP due to the fact that you are only using HP ProCurve switches.

Without seeing your configs, I would compare the below configs with yours.

Switch A
ip routing
interface 1
no lacp
interface 2
no lacp
interface 3
no lacp
interface 4
no lacp
trunk 1-4 Trk1 Trunk
vlan 1
tagged Trk1
ip address 1.1.1.1/24
exit
vlan 100
tagged Trk1
ip address 1.1.100.1/24
exit
vlan 200
tagged Trk1
ip address 1.1.200.1/24
exit

Switch B
ip default-gateway 1.1.1.1
interface 1
no lacp
interface 2
no lacp
interface 3
no lacp
interface 4
no lacp
interface 5
no lacp
interface 6
no lacp
trunk 1-4 Trk1 trunk
trunk 5-6 Trk2 trunk
vlan 1
tagged Trk1-Trk2
exit
vlan 100
tagged Trk1-Trk2
exit
vlan 200
tagged Trk1-Trk2
exit

Switch C
ip default-gateway 1.1.1.1
interface 1
no lacp
interface 2
no lacp
trunk 1-2 Trk1 Trunk
vlan 1
tagged Trk1
exit
vlan 100
tagged Trk1
exit
vlan 200
tagged Trk1
exit

[Matt Hobbs]

When configuring trunks, as Jody has suggested, it's best to configure them on both ends. You can use either 'trunk' or 'LACP' as the trunk type, as long as it's set on both ends.

With LACP configured as a static trunk on one end only, it will create a dynamic trunk on the other switch which can't be a member of multiple VLANs unless you configure GVRP (from memory).

[wtu]

Hi Jody,

Thank you for the prompt reply with the example and your valuable suggestion. Please refer to the attachment for the configurations on the three switches.

After I submitted the question, I found that my default gateway was specified wrong on one of the PCs used for testing. This caused the network connectivity to be lost. I used 192.168.20.3 and it should be always 192.168.20.1 for any devices connecting to VLAN 200. After I corrected the default gateway setting on that PC, the network connectivity was then restored.

There are two 3Com 3c17203 switches used for inter-building links using fibres. From one of the two switches, there is a port connecting to the Procurve 2848 switch which is currently in production. There are two Asante FM2900 switches used for another inter-building links. There exists a link between one of the FM2900 switches and the production 2848 switch. I'm configuring three new 2848 switches to upgrade the network. In order to avoid looping and accommodating switches only capable of doing 802.1d, could you make recommendations with examples? Although we don't have VLANs now, I would like to have VLANs configured to allow an upgrade in the near future.

Thanks!

[Matt Hobbs]

Do you have a network map that you could attached of what you'd what like to achieve?

From your configs, I'm assuming they're inteconnected via a few different ports, 33, 34, trunk1, trunk2.. if you could clarify this in the way of a network map it would really help.

When running with other devices that don't support 802.1w RSTP - you should set the interconnecting ports on the procurve to disable 'mcheck', and you should also change the port costs to match throughout the network. The spanning-tree chapter goes into more detail on this - ftp://ftp.hp.com/pub/networking/software/AdvTraff-Oct2005-59908853-Chap05-STP_RSTP.pdf

Most of the time though spanning-tree should figure itself out so you may not need to change anything at all, the above recommendations are really just that, recommendations.

[wtu]

Yes. They are interconnected with several trunk ports. Please refer to the simple drawing attached for the detail.

I've modified the configurations on the three new Procurve switches and used Procurve trunking but not LACP per Judy. Currently, there is only one Procurve 2848 switch with a flat subnet configured on it. I configured vlans across the three new switches with port trunking and would like to put these switches into production when the scheduled time comes. Then, I'll re-configure the cuurent production 2848. Additionally, I'll use SW1 as the new default gateway for servers, printers, and desktop computers (currently a Sonicwall firewall is being used as the default gateway). I'll also replace the Sonicwall with a CISCO ASA.

According to the current STP topology, the FM2900 SWA in building 3 is the root switch. FM2900 doesn't support RSTP. I'd like to use SW1 as the root switch instead.

Any ideas and/or recommendations? Thanks!

[Matt Hobbs · Accepted solution]

With your network map I can see your trunks between the 2848's. What I'd also do is have a single link going from the bottom 2848 back to the top 2848. That way if the middle switch dies, it's not going to completely isolate the one at the bottom.

There is a good chance though that the bottom switch may prefer this link to the trunk, so you may need to study spanning-tree priorities to make sure the links you wish to be blocked remain blocked in normal operation.

Also if you could include some extra links going to the second 2848 to the 3com and Cisco it would obviously help in the event that the top 2848 died.

This is all very paranoid of course, it's fairly rare that switches go faulty and if you can live with the possible downtime and move links manually in the event of a failure, then I would probably do that instead. (Keep it simple).

As for making your preferred swith the spanning-tree root, use the command 'spanning-tree priority 0'. That should do it, if not, you'll need to check the current root and give it a lower priority.

[jamesps]

Unless you want to monitor a trunk I would stay with LACP static. This gives you the advantage of customizing the STP parameters per trunk and also you can have the trunk operating in a different vlan other than the default one. This is extremely useful since you probably don't want to advertise the default vlan outside the switches own network. I would recommend creating vlans, IP them, dedicate management ports in that new vlan for management purposes then tag the ports accordingly. If you want to monitor a trunk then this option fails and you have to go with a "Trunk" type trunk.

In any case I would definetly leave all the ports with LACP ENABLED! This will detect, adjust and report any port/connectivity problem right away, LACP will block your port if something is wrong about it. Disabling LACP has no justification unless you want to avoid certain issues. It is a great feature you definetly want to use. If you decide to go with an LACP trunk or not that depends on your configuration and what you wish to achieve of course.

You can check the Management and Configuration guide there is a chapter which explains all about port trunking.

Hope this helps,
james

[wtu]

I really appreciate suggestions for establishing redundancies or HA by having additional links to the middle switch from the top and other switches located in other buildings. However, I decided to keep things simple at this stage without introducing potential spanning tree troubleshootings.

What I'd like to know is that if I run RSTP on the Procurve switches and STP is disabled while 3com and Asante FM2009 switches do not support RSTP, will I be able to use the command, spanning-tree priority 0, to force the top Procurve switch to act as the root bridge or I have to manipulate other paramenters on all switches to make it happen? What if the top switch acting as the default gateway fails, would there be a way for the middle or the bottom switch take over the default gateway IP originaaly assigned to VLAN 1 on the top switch?

Thanks!

[wtu]

I thought about creating another VLAN instead of advertising the default VLAN, VLAN 1, outside the switch's own network. However, if I do so, I have to create that VLAN on all other brand switches. They do not have layer 3 routing capabilities and are required to use an external router or a layer three switch for routing. In order to minimize modifications at this stage, I think it's better for me to live with a simple design for now and worry about others later on. I also thought about assigning VLAN 1 to every switch port as a tagged VLAN and then use it for port SPAN/mirroring purpose. Well, it could introduce too much change though.

I hesitate to make too much change now since I have been informed that, for some reasons, I won't be able to do console access to those switches located in other buildings at the time when I perform the network change because the change must be done outside regular hours and the buildings will not open then.

Does it make sense if I continue to use trunk type of trunks but create LACP type of trunks for redundancy in the future?

Thanks!

[jamesps]

About your first question, the answer is yes and no. If you want your "stack" of HP ProCurves to be redundant then you will need separate links to the Cisco (right) and to the 3Com (left). It is STP/RSTP/MSTP's job to block the uplinked port on the ProCurve if that link is redundant. Even if your Cisco or 3Com is not STP aware, the ProCurve switches will detect it since they will be in the same broadcast domain.
One issue is choosing which is the root switch, those links will have the highest priority and because they all have the same speed you will have to define each port/link priority manually. You can choose them as they are in the diagram, doesn't matter.
To make the whole setup redundant, make sure you tag all the vlans on the ports used for trunking.
Other than these issues, your setup will be fully redundant and pretty simple to configure as well.
Check the manual because there are some parameters you need to play with, you need to tell the switch basically not to send BPDUs on the links connected to the 3Com and CISCO, so they would not "complain" in any way. But that is for "fine tuning" if you experience any issues.
Are your 3Com and CISCO LACP capable? If yes then it will help, I would leave ALL the ports LACP enabled to make sure there are no loops "by mistake". With no STP enabled LACP will save you and disable the appropriate ports to avoid the loops!

Hope this helps,
james

[jamesps]

I would not do any STP change without having access to the switches. In certain cases you need to disable the ports while enabling STP on them. You will notice high activity on the switch after STP has been enabled and even lost connectivity to it.

Even if you add another vlan, IP it and want to use it for management purposes, you can simply UNTAG (and not TAG) the uplink ports to the Cisco and 3Com. The packets should be switched by the VLAN "unaware" devices at the other end just fine. If they would be tagged then certainly it won't work. The only reason I would use the DEFAULT_VLAN VID 1 would be tagging the trunk ports. For servers and everything else I would create separate vlans as needed. You could use a "chassis like" setup this way especially if you run out of ports on the HPs.

james

[wtu]

May I have some examples regarding path cost fine tuning for uplink ports? Could you explain what does "chassis like" setup? When I roll out the new Procurves onto the production, if I power off the CISCO and 3Com first and then power them back on, would the Procurve switch with the highest priority set (priority 0) takes over the root bridge position?

Thanks!

[jamesps]

Please see the txt file attached. The ports 36-38 are connected to end nodes (servers), 48 is connected to the main uplink (that would be your Cisco or 3Com) and Trk1 is a LACP dynamic trunk connected to an identical switch serving as a redundant backup. This other switch is also connected to the same uplink as the other one.

How the STP will route the traffic is set in the 4-th column which is the path COST. That means the links with the lowest number have the highest priority. This has nothing to do with the fact that this is my root switch or not.

The primary link is the one with path cost 20000 and the Trk1 is higher which means lower priority (my backup).
With these being said, STP's job is to block the Trk1 path while my port#48 is still active because there have to be no loops. The path costs is something you will have to come up with that depending on your network infrastructure and where do you want to send the traffic through.

The documentation about STP is a great help. I would recommend using MSTP even on 2800 series. I noticed a lower CPU load when using vlans intensively (as the manual says).

james

[wtu]

Should I enable STP on Procurves? I did a show spanning-tree and receive Rapid Spanning Tree (RSTP) Information saying STP Enabled: No

I am confused somewhat. I know RSTP has been enabled but does it co-exist with STP on the Procurves?

[jamesps]

A ProCurve 2800 supports: STP, RSTP and MSTP. The last one can be configured using CLI only. The others can be configured with the Menu as well. They all accomplish the same thing. If you have STP already in place then you can use path costs to establish a certain traffic priority. If you want these switches to talk to other "older" devices which cannot understand RSTP or MSTP, then you would have to stay with "basic" STP. In most cases even RSTP works ok. You would need to check those devices specifications and what STP versions do they support.

james

[jamesps]

If you want you can enable STP on the ProCurves and it will "talk" to the other STP aware devices in your network creating a loop-free network. If you will have redundant links to the ProCurves then this is a must. If not then you will probably not have a need for it. When enabling STP on the trunked ports you will lose connectivity with the switches for a few moments while the tree is "built". That is normal and I think the manual recommends disabling the ports before doing that.

james

[wtu]

If I use Lacp trunks, do I configure the trunk ports lacp active? If I want to enable STP on trunked ports, do I remove "no spanning-tree edge-port" accordingly?

I am revising my configurations to switch from trunk type of trunks to lacp type. Can I have examples showing how may I configure lacp type trunks to enable the detection/adjustment/report of connectivity issues?

Thanks.

[wtu]

I made a change on two trunks. When I did a "show lacp" one trunk is showing yes under lacp partner while the other showing no, how would I fix it?
Why that happened?