I read a lot of documentations regarding HQOS, but not sure how it's programmed on the C9300 ASIC UADP 2.0.
I have the following, before changing the configuration a Catalyst 9300 running a HQOS shaper on outbound traffic on uplink.
What I've got before :
service-policy CHILD
class CS4
bandwidth percent 20
[...]
+ 5 classes and the default BE
[...]
service-policy PARENT
class class-default
shape average 200000000
service-policy CHILD
When programming this one as output, we see each class is affected to one of the 8 queue available on C9300. Until classes are less than 7, no issues at all. But now I've 2 groups of each 5/6 DSCP.
What happen really if I create a new policy CHILD2 and change parent policy like this :
service-policy CHILD1
class CS1
bandwith percent 20
class CS2
bandwith percent 20
[...]
service-policy CHILD2
class AF11
bandwith percent 20
class AF21
bandwith percent 20
[...]
service-policy PARENT
class CHILD1
shape average 100000000
service-policy CHILD
class CHILD2
shape average 100000000
service-policy CHILD2
[...] + DEFAULT/BE
If I look at the "sh platform hardware fed switch 1 qos queue config interface gi1/0/2" command, I see only 2 ou 3 queues used. How does the PARENT apply on 8 queues ? How does the CHILD1 or CHILD2 policy-maps affect queues and bandwidth reservation for each class on each child policies ?
I'm not sure which traffic will be choose to be dropped when coming to shaping situation, for each DSCP classes ? Is it WRED ?