At 10:56 05/01/2011 +0100, Peter Rathlev wrote:
Content-Transfer-Encoding: 7bit
On Wed, 2011-01-05 at 11:36 +0200, Hank Nussbacher wrote:
> We have a 7613 w/ WS-SUP720-3BXL running 12.2(18)SXF11.
> We have a 48 port WS-X6748-GE-TX. On one interface we
> continue to see output drops when traffic goes above 200Mb/sec.
[snip]
Do you have QoS enabled? What does "show queueing interface Gi9/29" tell
you?
Output drops are "egress buffer overflow" drops, so technically it
happens because the box tries to send a packet out an interface already
in use (transmitting another packet) when there is no buffer space to
store the packet until transmission. Micro-bursts and oversubscription
are possible causes. Short queues exacerbate the problem. With "mls qos"
enabled and no interface-specific adjustments you could have (too) short
queues for the relevant traffic.
Also:
gp#sho mls qos
QoS is enabled globally
Policy marking depends on port_trust
QoS ip packet dscp rewrite enabled globally
Input mode for GRE Tunnel is Pipe mode
Input mode for MPLS is Pipe mode
Vlan or Portchannel(Multi-Earl) policies supported: Yes
Egress policies supported: Yes
----- Module [7] -----
QoS global counters:
Total packets: 3592763
IP shortcut packets: 0
Packets dropped by policing: 140
IP packets with TOS changed by policing: 96531
IP packets with COS changed by policing: 1700312
Non-IP packets with COS changed by policing: 0
MPLS packets with EXP changed by policing: 0
----- Module [9] -----
QoS global counters:
Total packets: 7865063
IP shortcut packets: 0
Packets dropped by policing: 765
IP packets with TOS changed by policing: 35559
IP packets with COS changed by policing: 8
Non-IP packets with COS changed by policing: 0
MPLS packets with EXP changed by policing: 0
-Hank
_______________________________________________
cisco-nsp mailing list [email protected]
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/
