Cisco 3850 | 16.6.8 | Total output drops

[Alan.A]

I have 3850 switch running in version 16.6.8 and few interfaces having packet drops, like connections going to access points.after a bit of research I come up with this configuration. Is this correct? This 3850 switches are use only for end users (wired, Phones and wireless (no servers).

conf t
!
qos queue-softmax-multiplier 1200
!
ip access-list extended ACL_QUEUE1
permit ip any any
!
class-map match-any CMAPQUEUE1
match access-group name ACL_QUEUE1
!
policy-map QUEUE1
class class-default
bandwidth percent 100

To be applied on interfaces going to Access Points and Uplink to Core
interface GigabitEthernetX/0/X
service-policy output QUEUE1

This is a sample of interface with AP connected

GigabitEthernet1/0/48 is up, line protocol is up (connected)
Hardware is Gigabit Ethernet, address is 00a5.bfdf.0b30 (bia 00a5.bfdf.0b30)
Description: ***AP8 ***
MTU 1500 bytes, BW 1000000 Kbit/sec, DLY 10 usec,
reliability 255/255, txload 1/255, rxload 1/255
Encapsulation ARPA, loopback not set
Keepalive set (10 sec)
Full-duplex, 1000Mb/s, media type is 10/100/1000BaseTX
input flow-control is on, output flow-control is unsupported
ARP type: ARPA, ARP Timeout 04:00:00
Last input never, output 00:00:00, output hang never
Last clearing of "show interface" counters 8w2d
Input queue: 0/2000/0/0 (size/max/drops/flushes); Total output drops: 20083437
Queueing strategy: fifo
Output queue: 0/40 (size/max)
5 minute input rate 2000 bits/sec, 2 packets/sec
5 minute output rate 2000 bits/sec, 2 packets/sec
303584396 packets input, 167930756495 bytes, 0 no buffer
Received 5931092 broadcasts (5219146 multicasts)
0 runts, 0 giants, 0 throttles
0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored
0 watchdog, 5219146 multicast, 0 pause input
0 input packets with dribble condition detected
680183326 packets output, 528204738590 bytes, 0 underruns
0 output errors, 0 collisions, 1 interface resets
0 unknown protocol drops
0 babbles, 0 late collision, 0 deferred
0 lost carrier, 0 no carrier, 0 pause output
0 output buffer failures, 0 output buffers swapped out

[MHM Cisco World]

No dont use it SW accept it but it dont have nay effect. 
see my other comment.

[MHM Cisco World]

That meaning I am right it 400 not 1200, 
2880 <<- softMax is increase  previous was 720

please still one point do you config ratio?

[Joseph W. Doherty]

Interesting.  Wonder what "qos queue-softmax-multiplier ?" shows?

[Alan.A]

sorry, its still "400"

XXXXXXXXXXXXXXXX#sh run | sec qos queue-softmax-multiplier
qos queue-softmax-multiplier 400
XXXXXXXXXXXXXXXX#show platform hardware fed switch active qos queue config interface gigabitEthernet 1/0/48
DATA Port:26 GPN:48 AFD:Disabled QoSMap:0 HW Queues: 208 - 215
DrainFast:Disabled PortSoftStart:3 - 4320
----------------------------------------------------------
DTS Hardmax Softmax PortSMin GlblSMin PortStEnd
----- -------- -------- -------- -------- ---------
0 1 5 120 7 1920 6 320 0 0 5 5760
1 1 4 0 8 2880 3 480 2 180 5 5760
2 1 4 0 5 0 5 0 0 0 5 5760
3 1 4 0 5 0 5 0 0 0 5 5760
4 1 4 0 5 0 5 0 0 0 5 5760

Let me change it to 1200.

[Alan.A]

This is the result.

XXXXXXXXXXXXX#show platform hardware fed switch active qos queue config interface gigabitEthernet 1/0/48
DATA Port:26 GPN:48 AFD:Disabled QoSMap:3 HW Queues: 208 - 215
DrainFast:Disabled PortSoftStart:2 - 10000
----------------------------------------------------------
DTS Hardmax Softmax PortSMin GlblSMin PortStEnd
----- -------- -------- -------- -------- ---------
0 1 4 0 6 10000 9 800 5 300 3 10000
1 1 4 0 5 0 5 0 0 0 3 10000
2 1 4 0 5 0 5 0 0 0 3 10000
3 1 4 0 5 0 5 0 0 0 3 10000
4 1 4 0 5 0 5 0 0 0 3 10000
5 1 4 0 5 0 5 0 0 0 3 10000
6 1 4 0 5 0 5 0 0 0 3 10000
7 1 4 0 5 0 5 0 0 0 3 10000
Priority Shaped/shared weight shaping_step
-------- ------------- ------ ------------
0 7 Shared 50 0
1 0 Shared 10000 0
2 0 Shared 10000 0
3 0 Shared 10000 0
4 0 Shared 10000 0
5 0 Shared 10000 0
6 0 Shared 10000 0
7 0 Shared 10000 0

Weight0 Max_Th0 Min_Th0 Weigth1 Max_Th1 Min_Th1 Weight2 Max_Th2 Min_Th2
------- ------- ------- ------- ------- ------- ------- ------- ------
0 0 7968 0 0 8906 0 0 10000 0
1 0 0 0 0 0 0 0 0 0
2 0 0 0 0 0 0 0 0 0
3 0 0 0 0 0 0 0 0 0
4 0 0 0 0 0 0 0 0 0
5 0 0 0 0 0 0 0 0 0
6 0 0 0 0 0 0 0 0 0
7 0 0 0 0 0 0 0 0 0

I have this error

XXXXXXXXXXXXX#policy-map QUEUE1
XXXXXXXXXXXXX#(config-pmap)# class class-default
XXXXXXXXXXXXX#(config-pmap-c)# bandwidth percent 100
XXXXXXXXXXXXX#(config-pmap-c)# priority level 1 percent 90 ?
XXXXXXXXXXXXX#(config-pmap-c)# priority level 1 percent 90
Priority not allowed in conjunction with bandwidth

[MHM Cisco World]

for me it seem priority level not work 
so as I mention before 
policy-map QUEUE1
class class-default
bandwidth percent 100
queue-buffers ratio 100 <<- WIHTOUT specify prioroty level.

please share the output of Queue.

[Joseph W. Doherty]

In the post I referenced in my reply to @Reza Sharifi, I too considered using PQ level 1, to take advantage of the hardbuffers.

However, within some 9K information (i.e. might not apply to 3850), it was also mentioned using level 1 would not take advantage of any soft buffer increase due to "qos queue-softmax-multiplier".  Further, that documentation noted, using PQ level  2, would use both hard buffers and take advantage of soft buffers increased by "qos queue-softmax-multiplier".  I.e. if only using 1Q, using PQ level 2 got the maximum benefit of hard buffers and soft buffers.

(BTW, what Cisco did makes sense, as they would recommend PQ level 1 for VoIP and PQ level 2 for real-time video.  The former is not, usually, as variable bandwidth demanding as the latter.)

Further, my understanding of the 9K documentation, if we're only going to define 1 Q, you might as well also provide it a 100% buffer ratio.  As PQ, though, no need to define a bandwidth percentage.

So, if you want to try using PQ, I would suggest seeing if device will accept:

qos queue-softmax-multiplier 1200 !or whatever maximum the device accepts

policy-map QUEUE1
class class-default
priority level 2
queue-buffers ratio 100

[MHM Cisco World]

his  SW with IOS XE give wrong Buffer size for both hard and soft. 
so I suggest in beginning to use ratio.
let see his reply.  

[Joseph W. Doherty]

Maybe a variation of what you posted, earlier?

I.e.

policy-map QUEUE1
class class-default
priority level 2 percent 100

[Joseph W. Doherty]

Disclaimer: I've vary familiar with the 3560/3750 QoS architecture, and just the other day, studied, a bit, on the Catalyst 9K QoS architecture.  I'm not as familiar with the 3650/3850 architecture, but believe it and the 9K are more similar that it's to the 3560/3750.  I.e. I may be inaccurate on some of what I suggest (hopefully not, though).

"qos queue-softmax-multiplier 1200"

Yea, that seems to be a generally safe and good change.

policy-map QUEUE1
class class-default
queue-buffers ratio 100 !with just 1 Q, something else to possibly add (also suggested by @MHM Cisco World )
bandwidth percent 100

The forgoing might be all you need to do.

See more, in my reply to policy-map suggested by @MHM Cisco World 

[Alan.A]

This is the config
***************************
!
qos queue-softmax-multiplier 1200
!
ip access-list extended ACL_QUEUE1
permit ip any any
!
class-map match-any CMAPQUEUE1
match access-group name ACL_QUEUE1
!
policy-map QUEUE1
class class-default
bandwidth percent 100
queue-buffers ratio 100

!To be applied on interfaces going to AP and Uplink to Core
interface GigabitEthernetX/0/X
service-policy output QUEUE1

***************************
This is the QUEUE

show platform hardware fed switch active qos queue config interface gigabitEthernet 1/0/48
DATA Port:26 GPN:48 AFD:Disabled QoSMap:0 HW Queues: 208 - 215
DrainFast:Disabled PortSoftStart:3 - 4320
----------------------------------------------------------
DTS Hardmax Softmax PortSMin GlblSMin PortStEnd
----- -------- -------- -------- -------- ---------
0 1 5 120 7 1920 6 320 0 0 5 5760
1 1 4 0 8 2880 3 480 2 180 5 5760
2 1 4 0 5 0 5 0 0 0 5 5760
3 1 4 0 5 0 5 0 0 0 5 5760
4 1 4 0 5 0 5 0 0 0 5 5760
5 1 4 0 5 0 5 0 0 0 5 5760
6 1 4 0 5 0 5 0 0 0 5 5760
7 1 4 0 5 0 5 0 0 0 5 5760
Priority Shaped/shared weight shaping_step
-------- ------------- ------ ------------
0 0 Shared 50 0
1 0 Shared 75 0
2 0 Shared 10000 0
3 0 Shared 10000 0
4 0 Shared 10000 0
5 0 Shared 10000 0
6 0 Shared 10000 0
7 0 Shared 10000 0

Weight0 Max_Th0 Min_Th0 Weigth1 Max_Th1 Min_Th1 Weight2 Max_Th2 Min_Th2
------- ------- ------- ------- ------- ------- ------- ------- ------
0 0 1625 0 0 1816 0 0 2040 0
1 0 2295 0 0 2565 0 0 2880 0
2 0 0 0 0 0 0 0 0 0
3 0 0 0 0 0 0 0 0 0
4 0 0 0 0 0 0 0 0 0
5 0 0 0 0 0 0 0 0 0
6 0 0 0 0 0 0 0 0 0
7 0 0 0 0 0 0 0 0 0

[MHM Cisco World]

that OK for me, even so I dont see any modify of queue after add ratio
but change the multiplier from 1200 to 400 (correct value) give more Queue depth and this can solve output drop, 

apply confing and check the drop for at least two days.
good luck friend

[MHM Cisco World]

that OK for me, even so I dont see any modify of queue after add ratio
but change the multiplier from 1200 to 400 (correct value) give more Queue depth and this can solve output drop, 

apply config and check the drop for at least two days.
good luck friend

[Alan.A]

Actually, I rolled back my change. I will test it first from LAB using 3650 switch. I will provide update once I applied it in production.

[MHM Cisco World]

keep update me 
thanks a lot