Sub-Interface Ping Loss on Wireless Connection

[eeebbunee]

Hello Professionals,

I have seen weird situation when ping to google dns and sub interface with my wireless connection.

When I ping from my pc to Google DNS (8.8.8.8) and sub-interface(internal network) at the same time,

  - Request Timed out of Google DNS : 1 or 2 times during 200 times ping.
     PC (10.10.10.50) --- Ping 8.8.8.8

  - Request Timed out of sub-interface: 8-10 times during 200 times ping.
     PC (10.10.10.50)  --- Ping 10.10.10.1

How is that possible? 

For some reason, during my PC failed to send ping to sub-interface for few seconds, ping to google dns were fine!

From Core switch, priority of sub-interface routing is higher than routing to last resort (default gateway). Isn't it?

My infra setup like this:

<<Core switch>> -- <<Wireless Controller 2504>> -- <<Cisco AP 2802I>> -- <<my PC>>

I'm confused with my result. Can it be the reason of wireless controller?

Please help me.

Can anyone guess why am I got this result?

[Rich R]

- Make sure your software version is up to date (links below).
- Isolate the problem by running packet captures on the switch ports (AP and core) so you at least know where the packet loss is happening (echo request lost or echo reply lost, where it's lost) then you know which device is responsible and can focus on that.

[Flavio Miranda]

Hi

  Rreply from google server and reply from the sub-interface  (you did not specify where this subinterface is, Core or WLC). are separated events. Ping reply can be delayed in the event of high CPU or high traffic as ICMP uses UDP protocol and have no priority whatsoever,

When you ping Google you may receive all the packets cause Google server is healthy enough and you may lose some packet due internet or your network problem. But most of packet will get responded.

When you ping your device either your WLC or Core, this device can be facing performance isse and will drop packet for ICMP but they are still able to allow traffic to/from internet.

 That´s could be one reason.

[Rich R · Accepted solution]

Minor correction - ping uses ICMP and ICMP does not use UDP - it is an IP protocol itself - protocol number 1, UDP is 17:
https://www.iana.org/assignments/protocol-numbers/protocol-numbers.xhtml

But the principle is correct.  The ping to core requires a process-switched reply from switch/router CPU so is subject to CPU load and control plane policing amongst other things while traffic to internet is simply switched (data plane).