[e-nsp] Problems using LACP-Trunks on X670 in a stack

[Chris Hellkvist]

Hi all,

i have two X670 (Rev 6.0 BootROM: 2.0.1.6    IMG: 15.1.2.12) running in a stack:
Slots    Type                 Configured           State       Ports
--------------------------------------------------------------------
Slot-1   X670-48x             X670-48x             Operational   48
Slot-2   X670-48x             X670-48x             Operational   48

The switches are used for 1/10G aggregation. The stack has a 20G (LACP
trunk on 1:1 and 1:2) uplink to a core router (Brocade MLXe box).
Different network segments are put in different VLANs, spanned over
the 20G trunk to the Brocade box that is terminating IP. The VLAn
config is quite easy and looks like this:
	Ports:	 5. 	  (Number of active ports=5)
           Untag:    *1:9,  *1:10,   *2:9,  *2:10
           Tag:       *1:1g

The other config is quite easy, nothing special or other features are
configured.
The untagged ports are connected to different switches, the tagged
load sharing ports are connected to the MLXe box.

So far, this config works quite fine without any problems with some 1G
and 10G ports.
The hassle starts when we are connecting 2 ports that are configured
as LACP trunk. As soon as these ports are physically connected there
is heavy packet loss on all ports connected to the X670 stack (not
100% loss, but around 80-90%). To resolve this situation the newly
connected ports have to be disconnected and the uplink ports facing to
the MLXe have to be disabled and enabled again. Everything other
connected to the MLXe box also works fine in the time the X670 stack
is producing loss

The logs of both the Extreme stack and the MLXe do not show any
problems (attached to the end of the mail)
The sharing config for the ports looks ok:
==============================================================================
  2:31               LACP      L2          2:31       -      R        0
                               L2          2:32       -      R        0
==============================================================================

 The device behind this trunk is a HP Procurve switch, the config of
the device is fine and also works when connected to other devices
(LACP is working and building the trunk group when used with a Cisco
or Brocade Box).

Anybody experienced similar problems?

Thanks,
Chris


For reference, the logs of the X670. The MLXe log does not show
anything in the time the problem occurs.

Slot-1 Marsellus-Wallace(EXSW-STACK).3 # sh log
02/20/2013 12:56:39.37 <Info:LACP.RemPortFromAggr> Slot-1: Remove port
2:30 from aggregator
02/20/2013 12:56:39.37 <Info:vlan.dbg.info> Slot-1: Port 2:30 is Down,
remove from aggregator 2:29
02/20/2013 12:56:39.37 <Info:vlan.msgs.portLinkStateDown> Slot-1: Port
2:30 link down - Local fault
02/20/2013 12:56:37.62 <Info:LACP.RemPortFromAggr> Slot-1: Remove port
2:29 from aggregator
02/20/2013 12:56:37.62 <Info:vlan.dbg.info> Slot-1: Port 2:29 is Down,
remove from aggregator 2:29
02/20/2013 12:56:37.62 <Info:vlan.msgs.portLinkStateDown> Slot-1: Port
2:29 link down - Local fault
02/20/2013 12:55:35.52 <Info:LACP.AddPortToAggr> Slot-1: Add port 2:30
to aggregator
02/20/2013 12:55:35.47 <Info:vlan.msgs.portLinkStateUp> Slot-1: Port
2:30 link UP at speed 1 Gbps and full-duplex
02/20/2013 12:55:17.13 <Info:LACP.AddPortToAggr> Slot-1: Add port 2:29
to aggregator
02/20/2013 12:55:15.09 <Info:vlan.msgs.portLinkStateUp> Slot-1: Port
2:29 link UP at speed 1 Gbps and full-duplex