07-16-2021 05:59 AM
I'm running test with new policy-maps on same hardware and same IOS version. Should be all the same configuration at the basis. When running test with the WRED configuration. Each class-map (INTERACT/TRANSACT/BULK) created also a 'default' queue.
Do not find the reason.. any idea?
show policy-map interface xx :
Class-map: CM_QOS_C_INTERACT (match-any) 435972 packets, 171050070 bytes 5 minute offered rate 239000 bps, drop rate 0000 bps Match: qos-group 31 0 packets, 0 bytes 5 minute rate 0 bps Match: qos-group 32 412960 packets, 168111830 bytes 5 minute rate 233000 bps Match: dscp af31 (26) af32 (28) af33 (30) 23012 packets, 2938240 bytes 5 minute rate 3000 bps Queueing queue limit 400 packets (queue depth/total drops/no-buffer drops) 3/71/0 (pkts output/bytes output) 431592/186310752 bandwidth remaining 60% Exp-weight-constant: 9 (1/512) Mean queue depth: 0 packets dscp Transmitted Random drop Tail drop Minimum Maximum Mark pkts/bytes pkts/bytes pkts/bytes thresh thresh prob default 8030/11407044 10/13380 61/86906 20 40 1/10 af31 85/12830 0/0 0/0 320 400 1/10 af32 423477/174890878 0/0 0/0 280 400 1/10 af33 0/0 0/0 0/0 240 400 1/10
And it does match traffic. The tagging is ok, as the correct service falls in the correct queue. However, why is the 'default' queue created?
07-16-2021 06:27 AM
Hello @_|brt.drml|_ ,
you have enabled WRED for this class-map and this may be the reason why a queue for DSCP default has been created.
However, the class-map should process only packets :
>>
Match: dscp af31 (26) af32 (28) af33 (30)
So there should not be any match in the DSCP default queue unless you are using also a policer that marks down non conforming packets.
Hope to help
Giuseppe
07-16-2021 08:11 AM
In your posting, there's no default queue, there's only one queue, the class queue.
When DSCP WRED is enabled, all the possible DSCP values have WRED default parameters (or at least that's how it used to be).
What might be happening here, only DSCP values that override the defaults and/or are seeing their corresponding DSCP values, might be shown with this command. (I recall [?] this command, on older equipment, showed all the DSCP WRED settings.)
What's interesting in your stats, DSCP BE is showing "hits", maybe matching in qos-group 32.
07-18-2021 11:29 PM
Just to repeat, each class has that 'default' queue enabled...
I'll have a look if there is a difference with the [?] as suggested commands between both systems.
The funny thing, items are correctly hit by the ACL, are placed in the correct queue. However, the 'default' queue in each class-map is also getting 'hits'.
Each class-map 'something-something' with this default queue has different 'stats'. So they are all different.
Weird stuff.
Thank you for the answers. If I find something, then I'll post it.
07-19-2021 08:18 AM
"Just to repeat, each class has that 'default' queue enabled... "
Again, also to repeat (for your class, as you posted [as there's no FQ is enabled in class]) there's only one queue per class (in non-LLQ policy classes).
What you're seeing is how WRED deals with BE marked packets.
Yes, there's the question of why you're getting hits on those. Assuming BE marked packets won't match any of the class rule's, there shouldn't be any in that class. Also again, since you didn't post all the supporting parts of the config for this policy, cannot say if that's happening. Also, possible, stats were from earlier version of class-map rules.
07-19-2021 10:34 PM
This is the Policy_Map. It has NV 'NO Video' expected on that line. Legacy stuff that is disappearing.
The 'Class DEFAULT' is exactly the same as the Cisco Default, also Legacy : Can be seen as the scavenger class.
policy-map PM_QOS_C_NV-85-5-0-10 class CM_QOS_C_VOICE priority level 1 police cir percent 9 class CM_QOS_C_ROUTING bandwidth remaining percent 2 class CM_QOS_C_NETMGT bandwidth remaining percent 2 set mpls experimental topmost 7 class CM_QOS_C_CALLSIG bandwidth remaining percent 2 set mpls experimental topmost 7 class CM_QOS_C_INTERACT bandwidth remaining percent 60 queue-limit 400 packets random-detect dscp-based random-detect dscp 26 320 400 10 random-detect dscp 28 280 400 10 random-detect dscp 30 240 400 10 class CM_QOS_C_TRANSACT bandwidth remaining percent 10 queue-limit 400 packets random-detect dscp-based random-detect dscp 18 320 400 10 random-detect dscp 20 280 400 10 random-detect dscp 22 240 400 10 class CM_QOS_C_BULK bandwidth remaining percent 8 queue-limit 400 packets random-detect dscp-based random-detect dscp 10 320 400 10 random-detect dscp 12 280 400 10 random-detect dscp 14 240 400 10 class CM_QOS_C_HIGHNET bandwidth remaining percent 5 random-detect random-detect precedence 0 25 64 10 random-detect precedence 1 30 64 10 random-detect precedence 2 35 64 10 random-detect precedence 3 40 64 10 random-detect precedence 4 45 64 10 random-detect precedence 5 50 64 10 random-detect precedence 6 55 64 10 random-detect precedence 7 60 64 10 class CM_QOS_C_PUBNET bandwidth remaining percent 10 random-detect random-detect precedence 0 25 64 10 random-detect precedence 1 30 64 10 random-detect precedence 2 35 64 10 random-detect precedence 3 40 64 10 random-detect precedence 4 45 64 10 random-detect precedence 5 50 64 10 random-detect precedence 6 55 64 10 random-detect precedence 7 60 64 10 class CM_QOS_C_DEFAULT bandwidth remaining percent 1 fair-queue queue-limit 1000 packets class class-default fair-queue random-detect random-detect precedence 0 60 64 1 police cir percent 1
The marker at the VLAN:
policy-map PM_QOS_M_CORPNET_D class CM_QOS_P_VOICE set ip dscp ef set mpls experimental imposition 5 set qos-group 50 class CM_QOS_P_VIDEO set ip dscp af41 set mpls experimental imposition 4 set qos-group 40 class CM_QOS_M_VOICE set ip dscp ef set mpls experimental imposition 5 set qos-group 50 class CM_QOS_P_NETMGT set ip dscp cs2 set mpls experimental imposition 7 set qos-group 29 class CM_QOS_P_BULK set ip dscp af11 set mpls experimental imposition 0 set qos-group 11 class CM_QOS_P_TRANSACT set ip dscp af21 set mpls experimental imposition 0 set qos-group 21 class CM_QOS_P_INTERACT set mpls experimental imposition 0 set ip dscp af32 set qos-group 32 class CM_QOS_P_PRIORITY set qos-group 31 set ip dscp af31 set mpls experimental imposition 0 class class-default set ip dscp default set mpls experimental imposition 0 set qos-group 0
Example of the marking 'match'-any classes:
class-map match-any CM_QOS_P_TRANSACT match protocol ms-lync match protocol smtp match protocol mapi match access-group name ACL_IP_DST_DMZ match access-group name ACL_IP_SRC_DMZ
On the outgoing Policy-Map the Default Class matches on this:
class-map match-any CM_QOS_C_DEFAULT description Class Map for Policing Identification of Default Unclassified Traffic match dscp default match mpls experimental topmost 0 match qos-group 0
If you need something else for the assessment, mail me. I can't understand this. As I have another more recent configuration on the same hardware. The only change is that I updated the 'Namings' according the Cisco QoS book.
I would expect that the BE fall in the class-map match any DEFAULT
07-19-2021 11:00 PM
Read over it,, indeed some extra context would be helpful.
The OUT policy map on the tunnel (That is the child of the parenet 'shaper')
policy-map PM_QOS_C_NV-85-5-0-10 class CM_QOS_C_VOICE priority level 1 police cir percent 9 class CM_QOS_C_ROUTING bandwidth remaining percent 2 class CM_QOS_C_NETMGT bandwidth remaining percent 2 set mpls experimental topmost 7 class CM_QOS_C_CALLSIG bandwidth remaining percent 2 set mpls experimental topmost 7 class CM_QOS_C_INTERACT bandwidth remaining percent 60 queue-limit 400 packets random-detect dscp-based random-detect dscp 26 320 400 10 random-detect dscp 28 280 400 10 random-detect dscp 30 240 400 10 class CM_QOS_C_TRANSACT bandwidth remaining percent 10 queue-limit 400 packets random-detect dscp-based random-detect dscp 18 320 400 10 random-detect dscp 20 280 400 10 random-detect dscp 22 240 400 10 class CM_QOS_C_BULK bandwidth remaining percent 8 queue-limit 400 packets random-detect dscp-based random-detect dscp 10 320 400 10 random-detect dscp 12 280 400 10 random-detect dscp 14 240 400 10 class CM_QOS_C_HIGHNET bandwidth remaining percent 5 random-detect random-detect precedence 0 25 64 10 random-detect precedence 1 30 64 10 random-detect precedence 2 35 64 10 random-detect precedence 3 40 64 10 random-detect precedence 4 45 64 10 random-detect precedence 5 50 64 10 random-detect precedence 6 55 64 10 random-detect precedence 7 60 64 10 class CM_QOS_C_PUBNET bandwidth remaining percent 10 random-detect random-detect precedence 0 25 64 10 random-detect precedence 1 30 64 10 random-detect precedence 2 35 64 10 random-detect precedence 3 40 64 10 random-detect precedence 4 45 64 10 random-detect precedence 5 50 64 10 random-detect precedence 6 55 64 10 random-detect precedence 7 60 64 10 class CM_QOS_C_DEFAULT bandwidth remaining percent 1 fair-queue queue-limit 1000 packets class class-default fair-queue random-detect random-detect precedence 0 60 64 1 police cir percent 1
The marking happens on the VLAN; it matches on standard protocols or ACL's.
if I do a 'show policy-map' then I can see the wred configuration in a table. This is exactly the same then with the other router.
#sh policy-map PM_QOS_C_NV-85-5-0-10 Policy Map PM_QOS_C_NV-85-5-0-10 Class CM_QOS_C_VOICE priority level 1 police cir percent 9 conform-action transmit exceed-action drop Class CM_QOS_C_ROUTING bandwidth remaining 2 (%) Class CM_QOS_C_NETMGT bandwidth remaining 2 (%) set mpls experimental topmost 7 Class CM_QOS_C_CALLSIG bandwidth remaining 2 (%) set mpls experimental topmost 7 Class CM_QOS_C_INTERACT bandwidth remaining 60 (%) queue-limit 400 packets packet-based wred, exponential weight 9 dscp min-threshold max-threshold mark-probablity ---------------------------------------------------------- af31 (26) 320 400 1/10 af32 (28) 280 400 1/10 af33 (30) 240 400 1/10 default (0) - - 1/10
So they all have that default queue inside the class....Except in this router it is shown inside the 'sh policy-map interface tunnel' as shown in the beginning. With packets inside it.
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