Cat9500 IPSLA Packet Classification

[JimWicks]

Hoping there's someone can help with a problem I have with packet classification for IPSLA traffic on C9500-24Y4C running version 17.6.4 where the egress QoS policy on Cat9500 running the IPSLA probe is not queuing these locally generated packets into the correct egress queue (no match counters on the "show policy-map interface" command).

With a simple back-2-back link between two devices below, the egress QoS policy on Cat9500-A is not matching any of the IPSLA UDP-jitter packets in the correct queue, only seeing them forwarded in class-default

Cat9500-A (IPSLA-Source) <------------> Cat9500-B(IPSLA-Responder)

If I  move the IPSLA probe config onto an ISR4K while keeping QoS config the same then I do see packet matches on the correct egress queue on the Cat9500-A facing towards Cat9500-B, so the setup below works OK.

ISR4K (IPSLA-Source) <----------> Cat9500-A <----------> Cat9500-B(IPSLA-Responder)

This suggests that the locally sourced IPSLA packets are not hitting the QoS classifier for the egress interface, even though other locally sourced SNMP, Syslog and SSH packets are getting queued correctly according to their DSCP markings, so the issue appears specific to IPSLA.

Does anyone know, is there any other CLI knobs that I need to tweak to enable the correct queuing of these IPSLA packets on the source and responder devices ??    Thanks.

[MHM Cisco World]

Session Presentation (ciscolive.com)

I think you need to check this link, it talk about CoPP and Queue use for each traffic type.
the Queue number is different between Cat9 and ISR 

[JimWicks]

Thanks, I'll take a look at the doc.   

Yep, I realize that Cat9K and ISR4K have different queuing mechanisms.  What I was trying to highlight by moving the IPSLA-Client onto the ISR4K is to show that the same IPSLA packet appears to be queued differently by the egress Cat9K QOS-policy depending upon whether that packet is sourced from an external device or generated locally by the Cat9500 control-plane (proves that the IPSLA config is good and the egress QOS policy is good in terms of matching DSCP, etc).