06-19-2016 12:56 AM - edited 03-08-2019 06:16 AM
I have faced a strange issue in my enterprise, Voice quality being dropped suddenly & data traffic was flowing smooth. When checked with with command show mls qos inteface Gi1/0/25 statistics it was confirmed packets marked with dscp 46 was dopped. I have enabled a prioirty queue to fix this issue for time being, awaiting more users to use voice calls on Monday. Now there are no drops aftre prioirty quequ added
But i am not sure why voice traffic got congested when interface utilisation is very low. Its a Gig Link & hardly we see 20-30Mbps of traffic on it. Can anyone explain in 3750 can we expect such voice traffic congestion when interface utilisation is low? Is this due to SRR?
Below are switch configs. QOS experts ur input please :)
******#show running-config int gigabitEthernet 1/0/25
Building configuration...
Current configuration : 381 bytes
!
interface GigabitEthernet1/0/25
description Uplink to ***** port 3/23
switchport trunk encapsulation dot1q
switchport trunk native vlan 33
switchport trunk allowed vlan 33,36,39,49,101-107,113,155,220,222,250,251,333
switchport trunk allowed vlan add 383,484,511
switchport mode trunk
load-interval 30
priority-queue out---added
udld port aggressive
mls qos trust dscp
end
****#show mls qos interface gigabitEthernet 1/0/25 statistics
GigabitEthernet1/0/25 (All statistics are in packets)
dscp: incoming
-------------------------------
0 - 4 : 103646660 0 0 0 0
5 - 9 : 0 0 0 56269 0
10 - 14 : 192 0 0 0 0
15 - 19 : 0 305 0 0 0
20 - 24 : 0 0 0 0 648490
25 - 29 : 0 2 0 0 0
30 - 34 : 0 0 1150 0 135342592
35 - 39 : 0 0 0 0 0
40 - 44 : 696706255 0 0 0 0
45 - 49 : 0 116 0 3573939 0
50 - 54 : 0 0 0 0 0
55 - 59 : 0 0 0 0 0
60 - 64 : 0 0 0 0
dscp: outgoing
-------------------------------
0 - 4 : 260479455 0 0 0 0
5 - 9 : 0 0 0 2 0
10 - 14 : 0 0 0 0 0
15 - 19 : 0 0 0 0 0
20 - 24 : 0 0 0 0 0
25 - 29 : 0 669456 0 0 0
30 - 34 : 0 0 0 0 0
35 - 39 : 0 0 0 0 0
40 - 44 : 0 0 0 0 0
45 - 49 : 0 171338 0 31479 0
50 - 54 : 0 0 0 0 0
55 - 59 : 0 9428 0 0 0
60 - 64 : 0 0 0 0
cos: incoming
-------------------------------
0 - 4 : 936506247 4 69191 544164 1150
5 - 7 : 0 3593694 1053095
cos: outgoing
-------------------------------
0 - 4 : 260666238 2 0 669456 0
5 - 7 : 0 202817 0
output queues enqueued:
queue: threshold1 threshold2 threshold3
-----------------------------------------------
queue 0: 0 0 696706256
queue 1: 1911 1102891 3472807
queue 2: 0 0 135345478
queue 3: 0 2 331509145
output queues dropped:-this was showing drops before
queue: threshold1 threshold2 threshold3
-----------------------------------------------
queue 0: 0 0 0
queue 1: 0 0 0
queue 2: 0 0 0
queue 3: 0 0 0
Policer: Inprofile: 0 OutofProfile: 0
Solved! Go to Solution.
06-19-2016 09:28 AM
Hi
On access switch, if QoS isn't enabled, then every packets use the full capacity. As soon as QoS is enabled, you may experience issues if misconfigured. The default SRR has a weight of 25 for each queues (4 queues in total).
Voice traffic is placed on queue 1 and not prioritize if priority-queue out is not enabled, that's why you experienced some packets drops.
On your port configuration, you may use srr-queue bandwidth share command.
The command looks like srr-queue bandwidth share 1 30 35 5 (the 4 numbers are for the 4 queues).
As soon as you configure as well priority-queue out, the 1st number that correspond to the 1st queue, could be 0, the switch know that it's the priority queue and traffic on it will be passed before the others.
There are several cisco documentation well done. two of them:
https://supportforums.cisco.com/blog/149936/output-drops-due-qos-296035603750-switches
https://supportforums.cisco.com/document/31581/egress-qos
https://reggle.wordpress.com/2013/05/14/qos-part-v-hardware-queues-on-35603750-switch-platform/
With those 3 documentations, you will be able to understand a little bit more how QoS is working on access switches.
Hope this is a little bit more clear.
PS: Please don't forget to rate and mark as correct answer if this solved your issue.
06-20-2016 05:54 AM
Disclaimer
The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.
Liability Disclaimer
In no event shall Author be liable for any damages wha2tsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.
Posting
Low average interface utilization does not preclude there being transient congestion. (BTW, very low utilization can be a sign of very bad congestion.)
To also expand upon what was mentioned by supportlan, 3750s don't have a lot of buffer RAM (2 MB per 24 copper ports and also for the uplink ports), and the way 3750 allocates buffer RAM, by default, often leads to queue drops when QoS is enabled.
You mentioned enabling PQ fixed your VoIP drops. That's certainly a possible solution, especially for traffic in the PQ, but are the remaining queues still dropping? If so, at an acceptable level?
You mentioned, in your later post, you might disable QoS. That's also a possible solution, and it will often mitigate drops, but it leaves your VoIP traffic unprotected by QoS.
Perhaps the "best" long term solution, is to implement a QoS policy that meets your needs; taking best advantage of your equipment's capabilities.
On 3750s, I have had much success with changing buffer settings to mimic a common/shaped pool architecture. This works well when not all interface are busily transmitting queued traffic, but a few ports are dealing with a transient burst.
06-19-2016 09:28 AM
Hi
On access switch, if QoS isn't enabled, then every packets use the full capacity. As soon as QoS is enabled, you may experience issues if misconfigured. The default SRR has a weight of 25 for each queues (4 queues in total).
Voice traffic is placed on queue 1 and not prioritize if priority-queue out is not enabled, that's why you experienced some packets drops.
On your port configuration, you may use srr-queue bandwidth share command.
The command looks like srr-queue bandwidth share 1 30 35 5 (the 4 numbers are for the 4 queues).
As soon as you configure as well priority-queue out, the 1st number that correspond to the 1st queue, could be 0, the switch know that it's the priority queue and traffic on it will be passed before the others.
There are several cisco documentation well done. two of them:
https://supportforums.cisco.com/blog/149936/output-drops-due-qos-296035603750-switches
https://supportforums.cisco.com/document/31581/egress-qos
https://reggle.wordpress.com/2013/05/14/qos-part-v-hardware-queues-on-35603750-switch-platform/
With those 3 documentations, you will be able to understand a little bit more how QoS is working on access switches.
Hope this is a little bit more clear.
PS: Please don't forget to rate and mark as correct answer if this solved your issue.
06-19-2016 10:53 PM
Yes, yes your are right. I will go ahead with srr shapig/sharing modification. We dont need QOS here as hardly we have <100Mbps traffic & its a Gig link. Probably i will remove a qos.
Thanks very much as it was very interesting & new learning for me.Good day :)
06-20-2016 04:49 AM
If you are sure that you'll never had congestion on that link then don't configure qos and the switch will react simply.
If you want to manage QoS then you need to configure it correctly (srr shape/sharing, priority-queue, mapping) otherwise you will face big issues.
Thanks
06-20-2016 10:47 AM
One thing which i found the interface is quieng as below. Why 1 Gbps port showing operational BW to 100 Mbps? Is this default behavior if so Queue1 was getting 100/25=4 Mbps traffic which was th obvious cause. Can we change port bw limit to 1000 so that i get 40Mbps on Queue 1...just curious to know
****#show mls qos interface gigabitEthernet 1/0/25 queueing
GigabitEthernet1/0/25
Egress Priority Queue : enabled
Shaped queue weights (absolute) : 25 0 0 0
Shared queue weights : 25 25 25 25
The port bandwidth limit : 100 (Operational Bandwidth:100.0)
The port is mapped to qset : 1
!
06-20-2016 12:55 PM
The first you work with Catalyst 3750 switches, it could be confusing.
The port bandwidth limit is expressed in percent and not Mbps. This value is changeable by using the command : srr-queue bandwidth limit
I recommend to read the documentation from Cisco.
Below a command reference talking about srr-queue bandwidth limit:
http://www.cisco.com/c/en/us/td/docs/switches/lan/catalyst3750/software/release/12-2_55_se/commmand/reference/3750cr/cli3.html
There is a good doc on QoS for access switches, I recommend to any guy who start with mls qos:
http://www.unifiedguru.com/498/
Thanks
PS: Please don't forget to rate and mark as correct answer if this solved your issue
06-20-2016 05:54 AM
Disclaimer
The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.
Liability Disclaimer
In no event shall Author be liable for any damages wha2tsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.
Posting
Low average interface utilization does not preclude there being transient congestion. (BTW, very low utilization can be a sign of very bad congestion.)
To also expand upon what was mentioned by supportlan, 3750s don't have a lot of buffer RAM (2 MB per 24 copper ports and also for the uplink ports), and the way 3750 allocates buffer RAM, by default, often leads to queue drops when QoS is enabled.
You mentioned enabling PQ fixed your VoIP drops. That's certainly a possible solution, especially for traffic in the PQ, but are the remaining queues still dropping? If so, at an acceptable level?
You mentioned, in your later post, you might disable QoS. That's also a possible solution, and it will often mitigate drops, but it leaves your VoIP traffic unprotected by QoS.
Perhaps the "best" long term solution, is to implement a QoS policy that meets your needs; taking best advantage of your equipment's capabilities.
On 3750s, I have had much success with changing buffer settings to mimic a common/shaped pool architecture. This works well when not all interface are busily transmitting queued traffic, but a few ports are dealing with a transient burst.
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