LACP lag issue between switches and clients on same subnet.

[Charlie Daffarn]

hi all,

We have a new Windows 2012 R2 files server which will require maximum speed
comms between them and the users. Windows 2012 now allows, natively, for a
2+ nic team, using a dynamic LACP lag to the switch.

The server is in a specific vlan (1010), with the 2 x nic ports, and the
lag in that vlan, untagged. All client desktops are in different vlans,
based on their physical location (eg floor).

Once configured on the server, the switch config shows:

#show lacp

Aggregator: lag.0.12
Actor Partner
System Identifier: 00:01:f4:7e:89:bf d8:9d:67:79:12:94
System Priority: 32768 0
Admin Key: 32768
Oper Key: 32768 0
Attached Ports: ge.2.18-19

#show config vlan

*set vlan egress 1010
lag.0.12*;ge.1.1-7,26,29*;ge.2*.1-7,*18-19*,36;ge.3.1-7,17,19
*untagged*

#show port vlan ge.2.18-19;lag.0.12
ge.2.18 is set to 1010
ge.2.19 is set to 1010
lag.0.12 is set to 1010

#show port status ge.2.18-19;lag.0.12
Port Alias Oper Admin Speed Duplex Type
(truncated) Status Status (bps)
------------ ---------------- -------- ------- ------ -------
------------------
ge.2.18 dormant up 1.0G full 1000-t
rj45
ge.2.19 dormant up 1.0G full 1000-t
rj45
lag.0.12 up up 2.0G lag
3 of 3 ports displayed, 3 port(s) with oper status 'up' or 'dormant'.


Test copying a file to or from a client and the server on one particular
subnet, and another client on a different subnet, either on the same edge
switch, or a different switch, to the server, allows each client to each
use one of the connected nics, to achieve > 2gbps transfer speed.

If we copy from 2 clients, on the same subnet, the copy drops to half speed
on both client, ie, they both use the same connected nic. This behaviour
occurs copying to, or from the server.

In theory, the subnet the clients are on shouldn't affect the load balanced
lag, but it seems this is not the case.

The windows lacp config allows either a Dynamic Load Balanced Mode, or
Address Hash mode. We've tried both....both form the dynamic lag on the
switch, with the same copy speed results.

Is there any reason, from a network pov, you can think of why we're seeing
this behaviour...have you seen this before, and do you have any ideas on
how we can resolve it, save from having each user/client in it's own
subnet/vlan, which would lead to it's own, more significant admin related
problems!!

thanks all,
Charlie.
--
This e-mail is intended only for the named person or entity to which it is addressed and
contains valuable business information that is proprietary, privileged, confidential and/or
otherwise protected from disclosure. If you received this e-mail in error, any review, use,
dissemination, distribution or copying of this e-mail is strictly prohibited. Please notify
us immediately of the error via e-mail to ***@uk-group.net and please delete
the e-mail from your system, retaining no copies in any media. We appreciate your cooperation.

---
To unsubscribe from enterasys, send email to ***@unc.edu with the body: unsubscribe enterasys gneu-***@gmane.org

[[e-mail]]

Hola buen día,

Estaré fuera de la oficina del Lunes 7 al viernes 18 de abril, tendré acceso limitado a mis correos de voz y electrónicos por lo que demorare en mi respuesta.
Si tu asunto requiere atención inmediata, por favor comunicarse con David Aguilar al teléfono 9000 1777, al celular 55 5217 3678 o al correo ***@netcontroll.com.

Saludos,
Gracias.


---
To unsubscribe from enterasys, send email to ***@unc.edu with the body: unsubscribe enterasys gneu-***@gmane.org

[Erik Auerswald]

Hi,

the LAG obviously forms and the fact that routed connections use different
physical links shows that L3 information is used for hashing, not just L2.
Unless per frame round-robin is used, any combination of client and server
will always use the same physical link. Both directions use the same link.

My interpretation is that your test clients in the same subnet by chance
use the same link. To verify change the IP address of one of the clients
to get it to use the other physical link.

On an S-Series with current firmware you can use 'show lacp
outputportAlgorithm' to verify that the default of L3 (dip-sip) is used.
You can even change it with 'set lacp outportAlgorithm', but it do not
recommend this.

With a significant number of clients all links of the LAG will be used.

HTH,
Erik
--
Dipl.-Inform. Erik Auerswald http://www.fg-networking.de/
***@fg-networking.de T:[phone] M:[phone]

Gesellschaft für Fundamental Generic Networking mbH
Geschäftsführung: Volker Bauer, Jörg Mayer
Gerichtsstand: Amtsgericht Kaiserslautern - HRB: 3630

Post by Charlie Daffarn
hi all,
We have a new Windows 2012 R2 files server which will require maximum speed
comms between them and the users. Windows 2012 now allows, natively, for a
2+ nic team, using a dynamic LACP lag to the switch.
The server is in a specific vlan (1010), with the 2 x nic ports, and the
lag in that vlan, untagged. All client desktops are in different vlans,
based on their physical location (eg floor).
#show lacp
Aggregator: lag.0.12
Actor Partner
System Identifier: 00:01:f4:7e:89:bf d8:9d:67:79:12:94
System Priority: 32768 0
Admin Key: 32768
Oper Key: 32768 0
Attached Ports: ge.2.18-19
#show config vlan
*set vlan egress 1010
lag.0.12*;ge.1.1-7,26,29*;ge.2*.1-7,*18-19*,36;ge.3.1-7,17,19
*untagged*
#show port vlan ge.2.18-19;lag.0.12
ge.2.18 is set to 1010
ge.2.19 is set to 1010
lag.0.12 is set to 1010
#show port status ge.2.18-19;lag.0.12
Port Alias Oper Admin Speed Duplex Type
(truncated) Status Status (bps)
------------ ---------------- -------- ------- ------ -------
------------------
ge.2.18 dormant up 1.0G full 1000-t
rj45
ge.2.19 dormant up 1.0G full 1000-t
rj45
lag.0.12 up up 2.0G lag
3 of 3 ports displayed, 3 port(s) with oper status 'up' or 'dormant'.
Test copying a file to or from a client and the server on one particular
subnet, and another client on a different subnet, either on the same edge
switch, or a different switch, to the server, allows each client to each
use one of the connected nics, to achieve > 2gbps transfer speed.
If we copy from 2 clients, on the same subnet, the copy drops to half speed
on both client, ie, they both use the same connected nic. This behaviour
occurs copying to, or from the server.
In theory, the subnet the clients are on shouldn't affect the load balanced
lag, but it seems this is not the case.
The windows lacp config allows either a Dynamic Load Balanced Mode, or
Address Hash mode. We've tried both....both form the dynamic lag on the
switch, with the same copy speed results.
Is there any reason, from a network pov, you can think of why we're seeing
this behaviour...have you seen this before, and do you have any ideas on
how we can resolve it, save from having each user/client in it's own
subnet/vlan, which would lead to it's own, more significant admin related
problems!!
thanks all,
Charlie.
--
This e-mail is intended only for the named person or entity to which it is addressed and
contains valuable business information that is proprietary, privileged, confidential and/or
otherwise protected from disclosure. If you received this e-mail in error, any review, use,
dissemination, distribution or copying of this e-mail is strictly prohibited. Please notify
the e-mail from your system, retaining no copies in any media. We appreciate your cooperation.
---

---
To unsubscribe from enterasys, send email to ***@unc.edu with the body: unsubscribe enterasys gneu-***@gmane.org