HP2810's LACP Trunking Help

[andreasj64]

I just bought 3 HP2810's, and wanted to connect each of them together with 2 cables using LACP.  I went into the terminal menu and made 2 of the ports Trk1 LACP on 2 of the switches and networked them together.  Seemed to work, but I'm not exactly sure if that's all I need to do.  I was also unclear on whether I should use spanning tree or not.  I also don't understand if there's any benefit to switching on Flowcontrol.  

I'm a newbie, so any good layman explanation or instruction would be appreciated.  I have a total of 4 hp2810 switches (3-48's and 1-24).  I want them to communicate at 2gb. 

When I run the "show lacp" command, everything looks okay to me.  All say active, up, success, etc. 

I am also using stacking for convenience.

I think it's right, but I tested a large file transfer across the trunk link, and I only saw activity on one of the ports within the web interface.  Shouldn't I have seen the load balanced across the 2 trunk ports or does it pick a single route/port depending on traffic?  

Just want to make sure I'm connecting them correctly before I install the 3 new ones in production. 

[spice2003]

Hi

I didn't see your full running config but it seems like you did ok, LACP in procruve is quit simple.

for example:

switch01(config)#trunk 1,2 trk1 lacp

switch02(config)#trunk 1,2 trk1 lacp

About the traffic on one interface, aggreagtin of few port's, doesnt mean that the traffic from point to point will spread out across the aggregation ports, if one session was started on port number 1, the entire session will be on port number-1.

A lot of pepole are thinking that in trunk mode, if one interface is full, the traffic will go thogurh the other interface, no it's not, a packet loss will occured.

The load balancing mechanisim is based on MAC address if im right, so if everything is configured ok, traffic on 1 link it's ok from time to time..

I hope I helped you

[andreasj64]

I just got my 3rd switch in, the 2810-24g.

I had to turn on spanning tree because my test were creating a broadcast storm throughout the network.  I took everyone down momentarily...  lol  touchy touchy.  I'm not sure if I'll need it when I go live, but my test environment runs through a little switch on my desk.  Both my computer and the main network connection run through that little cheapo unmanaged switch.  I think that's probably what's causing the issue, but I'm not certain.

Does spanning tree kill the LACP, or is the switch smart enough to know the ports are merged?

Is Flowcontrol necessary? 

Here's the "show run" info from all 3 of my switches, and the show LACP follow at the very end.

******************************************************************ProCurve48-A# show run

Running configuration:

; J9022A Configuration Editor; Created on release #N.11.64

hostname "ProCurve48-A"
snmp-server contact "Admin"
snmp-server location "Computer Room"
interface 46
   flow-control
exit
interface 47
   flow-control
exit
interface 48
   flow-control
exit
interface 45
   flow-control
exit
trunk 46,48 Trk1 LACP
trunk 45,47 Trk2 LACP
snmp-server community "public" Unrestricted
vlan 1
   name "DEFAULT_VLAN"
   untagged 1-44,Trk1-Trk2
   ip address 192.168.2.233 255.255.255.0
   exit
fault-finder bad-driver sensitivity high
fault-finder bad-transceiver sensitivity high
fault-finder bad-cable sensitivity high
fault-finder too-long-cable sensitivity high
fault-finder over-bandwidth sensitivity high
fault-finder broadcast-storm sensitivity high
fault-finder loss-of-link sensitivity high
fault-finder duplex-mismatch-HDx sensitivity high
fault-finder duplex-mismatch-FDx sensitivity high
stack commander "PPCStack"
stack member 1 mac-address f06281921700
stack member 2 mac-address 001db3bbf940
spanning-tree
spanning-tree Trk1 priority 4
spanning-tree Trk2 priority 4

************************************************************************ProCurve48-B# show run

Running configuration:

; J9022A Configuration Editor; Created on release #N.11.64

hostname "ProCurve48-B"
snmp-server contact "Admin"
snmp-server location "Computer Room"
interface 45
   flow-control
exit
interface 47
   flow-control
exit
trunk 45,47 Trk1 LACP
snmp-server community "public" Unrestricted
vlan 1
   name "DEFAULT_VLAN"
   untagged 1-44,46,48-Trk1
   ip address 192.168.2.234 255.255.255.0
   exit
fault-finder bad-driver sensitivity high
fault-finder bad-transceiver sensitivity high
fault-finder bad-cable sensitivity high
fault-finder too-long-cable sensitivity high
fault-finder over-bandwidth sensitivity high
fault-finder broadcast-storm sensitivity high
fault-finder loss-of-link sensitivity high
fault-finder duplex-mismatch-HDx sensitivity high
fault-finder duplex-mismatch-FDx sensitivity high
stack join f0628192f200
spanning-tree
spanning-tree Trk1 priority 4

**************************************************************Procurve24-A# show run

Running configuration:

; J9021A Configuration Editor; Created on release #N.11.64

hostname "Procurve24-A"
snmp-server contact "Admin"
snmp-server location "Computer Room"
interface 22
   flow-control
exit
interface 24
   flow-control
exit
trunk 22,24 Trk2 LACP
snmp-server community "public" Unrestricted
vlan 1
   name "DEFAULT_VLAN"
   untagged 1-21,23,Trk2
   ip address 192.168.2.236 255.255.255.0
   exit
stack join f0628192f200
spanning-tree
spanning-tree Trk2 priority 4

****************************************************************Procurve24-A# show lacp

                           LACP

   PORT   LACP      TRUNK     PORT      LACP      LACP
   NUMB   ENABLED   GROUP     STATUS    PARTNER   STATUS
   ----   -------   -------   -------   -------   -------
   22     Active    Trk2      Up        Yes       Success
   24     Active    Trk2      Up        Yes       Success

*********************************************************ProCurve48-B# show lacp

                           LACP

   PORT   LACP      TRUNK     PORT      LACP      LACP
   NUMB   ENABLED   GROUP     STATUS    PARTNER   STATUS
   ----   -------   -------   -------   -------   -------
   45     Active    Trk1      Up        Yes       Success
   47     Active    Trk1      Up        Yes       Success

*****************************************************ProCurve48-A# show lacp

                           LACP

   PORT   LACP      TRUNK     PORT      LACP      LACP
   NUMB   ENABLED   GROUP     STATUS    PARTNER   STATUS
   ----   -------   -------   -------   -------   -------
   45     Active    Trk2      Up        Yes       Success
   46     Active    Trk1      Up        Yes       Success
   47     Active    Trk2      Up        Yes       Success
   48     Active    Trk1      Up        Yes       Success

[andreasj64]

"

About the traffic on one interface, aggreagtin of few port's, doesnt mean that the traffic from point to point will spread out across the aggregation ports, if one session was started on port number 1, the entire session will be on port number-1.

A lot of pepole are thinking that in trunk mode, if one interface is full, the traffic will go thogurh the other interface, no it's not, a packet loss will occured.

The load balancing mechanisim is based on MAC address if im right, so if everything is configured ok, traffic on 1 link it's ok from time to time..

"

I guess I'm a little confused.  I thought the bandwidth doubled in my scenario, because I'm combining the two ports in LACP.  The only way that could be possible from what I can imagine is to send the data down both routes simultaneously and recombine the data on the other end as one single stream.  I don't understand enough about the protocol to know exactly what I'm accomplishing I suppose, other than redundancy at this point.  Is there truly a performance benefit on this particular switch? 

[paulgear]

Hi andreasj64,

As spice2003 mentioned, the spreading of traffic across your LACP trunks is done in a deterministic manner, by default using MAC addresses (on some switches you can tune this to something different). So all traffic between a pair of nodes will always follow the same path unless it is down.

So you are doubling your bandwidth, but you can't necessarily use it all because of the way load balancing works.

Brocade is the only vendor I know of who offers automated trunking of ports on a per-packet basis, and that is only on their latest range of switches. Details here: http://packetpushers.net/pq-show-11-brocade-vdx-8770-technical-deep-dive/

[andreasj64]

Are you saying that the path between paired trunks is somewhat random, so somtimes the stream will flow from switchA to SwitchB using ChannelA, and then sometimes it will flow throw ChannelB?

Just to clarify in a way I can understand.

SwitchA:

Paired trk1 LACP trunk on Ports 46 & 48

SwitchB

Paired trk1 LACP trunk on Ports 45 & 47

ChannelA = 46 & 45

ChannelB = 48 & 47

Although the LACP trunk has 2GB available, performance wise a single stream between hosts is never exceeding 1GB of performance?

Just want to make sure I understand it fully.  I'm new to switching.

I've also seen dyn-1, dyn-2 on some ports on another router which was installed well before my time.  Those ports lead to a specialized server which handles very large print files.  What is that anyway?  I'm assuming it means dynamic, but not sure of the difference between trunk and dynamic.  Is one better than the other?

[Vince_Whirlwind]

---

@andreasj64 wrote:

Are you saying that the path between paired trunks is somewhat random, so somtimes the stream will flow from switchA to SwitchB using ChannelA, and then sometimes it will flow throw ChannelB?

---

That's not it - each pair of hosts that communicate with each other will communicate via one of the "channels" in your LACP.

The switch determines the forwarding path within a Link Aggregation Group by using the Link Aggregation hash algorithm. This hash takes can take various factors into account (and is sometimes configurable on some devices): 
- source/destination MAC addresses 
- EtherType 
- IP source/destination addresses 
- IP protocol number 
- TCP/UDP source/destination ports

Generally, it's just an "IP hash" - the switch has two paths to choose from, it looks at the destination IP address, if it's an odd number it chooses path1, if it's an even number it chooses path2.

Therefore the same destination IP address will always use the same physical channel.