01-23-2012 09:39 AM
I have been reading the nk1v QoS guide and the fair queuing white paper which show quite clearly how to use the in build nk1 and vmw protocol matches to apply bandwidth guarantees to the uplink (egress) – However, I would like to assign cos markings for the upstream N5Ks.
Can I create a policy which would be assigned to each port profile to set cos, then use these markings on the CWFQ policy on the uplink profile? This way I assign bandwidth policy based on cos on the uplink as well as looking for the markings to do the same on the n5k... If so do you have an example of this config?
Solved! Go to Solution.
02-13-2012 09:10 AM
Troubleshooting this issue I used the command "show system internal ipqos event-history errors" which stated that I was not providing bandwidth for n1k_control, mgmt, packet.
I removed the policy-map which was marking cos 6 from the veth port-profile (all N1k data is marked with cos 6 by default so I did not need this anyway) - I also removed the class-map relating to this then created the class-map based on the n1k_control, mgmt, packet protocol, lastly, I referenced this within the uplink queuing policy which seems to have done the trick.
01-23-2012 10:09 AM
Check out this link. It shows you how to set the marking on the N1KV side.
http://www.vnephos.com/index.php/2011/11/another-nexus-1000v-wfq-qos-example/
louis
01-23-2012 10:16 AM
Hi - I have seen this link, however, it is not clear where the policy should be applied. I have added (above) where I think policy should be applied.
01-23-2012 10:11 AM
Below is an example I have put together from the various documents - Does it look correct?? (it only includes NFS for now)
n1kv# show policy-map
Type qos policy-maps
====================
policy-map type qos class-cos4
class class-default
set cos 4
Type queuing policy-maps
========================
policy-map type queuing nfs-policy
class type queuing nfs-class
bandwidth percent 30
nav-n1kv# show class-map
Type queuing class-maps
========================
class-map type queuing match-all nfs-class
match cos 4
port-profile nfs
type: Vethernet
description:
status: enabled
max-ports: 32
min-ports: 1
inherit:
config attributes:
switchport mode access
switchport access vlan 892
service-policy input class-cos4
no shutdown
evaluated config attributes:
switchport mode access
switchport access vlan 892
service-policy input class-cos4
no shutdown
assigned interfaces:
Vethernet10
port-group: nfs
system vlans: 892
capability l3control: no
capability iscsi-multipath: no
port-profile role: none
port-binding: static
port-profile uplink
type: Ethernet
description:
status: enabled
max-ports: 32
min-ports: 1
inherit:
config attributes:
switchport mode trunk
switchport trunk allowed vlan 892,900,998
service-policy type queuing output nfs-policy
channel-group auto mode on mac-pinning
no shutdown
evaluated config attributes:
switchport mode trunk
switchport trunk allowed vlan 892,900,998
service-policy type queuing output nfs-policy
channel-group auto mode on mac-pinning
no shutdown
assigned interfaces:
port-channel1
Ethernet3/7
Ethernet3/8
port-group: uplink
system vlans: 892,900,998
capability l3control: no
capability iscsi-multipath: no
port-profile role: none
port-binding: static
01-23-2012 07:40 PM
Hi Nael. So I'm a little confused on what you are asking. Are you asking to mark a CoS value for virtual machines that can match up to an upstream switch (i.e. Nexus 5000/5500 or 7000)? May I ask what you are trying to achieve?
01-24-2012 12:35 AM
Hi
I am looking to mark the different traffic classes attached to the different port profiles (nfs, vmotion, vm data, mgmt etc). This is so I can apply bandwidth guarantees on the upstream N5K as well as on the N1K. Once marked I can apply CBWFQ on the uplink port profile (N1K) using the COS markings I applied on ingress from the vethernet ports.
If I mark at the N1K the qos configuration on the N5K becomes much simpler as opposed to having to create ACLs to classify the traffic.
02-09-2012 12:35 PM
The above configuration does not seem to work. If I run "module vem 3 execute vemcmd pinst 305" I can see the correct number of queues, however there are no matches whatsoever. This makes me think that the cos value is not being applied.
I have seen a few blogs / documents stating a similar config, however, has anyone actually checked to see if the queuing policy on the uplink profile is matching?
This may have something to do with cos only being applied on dot1q trunk (which the uplink actually is) or, does the cos value need to be trusted on egress?
Does anyone have any insight into this? Or can someone with a similar config check to see if the classes are being matched on the uplink?
02-13-2012 09:10 AM
Troubleshooting this issue I used the command "show system internal ipqos event-history errors" which stated that I was not providing bandwidth for n1k_control, mgmt, packet.
I removed the policy-map which was marking cos 6 from the veth port-profile (all N1k data is marked with cos 6 by default so I did not need this anyway) - I also removed the class-map relating to this then created the class-map based on the n1k_control, mgmt, packet protocol, lastly, I referenced this within the uplink queuing policy which seems to have done the trick.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide