How to check drops on PA-500 due to performance limits

[VAntonenko]

Hello!

Is there way to check session/packet drops on PA-500 due to performance limitaions?

Scenaro:

There is PA-500 appliance, that provide internet access to +2000 users with agentless integration with AD.

Management CPU average load is about 70-100 %

Data plane CPU: 40-60%

Session utilization 40-60%

Main internet-access rule has Threat-Prevention profiles, and some rules without profiles for isolated guest Internet-access. 

Average used throughput is about 100-250mbps.

In datasheet PA-500 maximum throughput 250mbps and 100mbps with Threat-Prevention.

So it works near to perfomance limits, the question is about counters, or troubleshooting commands that can indicate service impacts resulting from PA-500 limits.

For example on FWSM there is a command: show np block, that show counters of packets dropped by network processor.

On this PA-500 i noticed increasing of:

packets dropped by flow state check in show interface command

and

flow_tcp_non_syn
flow_tcp_non_syn_drop
tcp_syn_missing

in show counter global command

P.S.

Is there command to check Accelerated Aging events?

Regards.

[reaper · Accepted solution]

hi !

The counters you mention are indicators of malformed sessions, where an ack packet is received without having first seen a syn. This usually only happens if either there is asymmetrical routing, an ack packet takes so long to arrive the session has already been deconstructed or there is an injection attack underway

there is no one counter to indicate packets dropped packets due to exceeding the system limitations as we'll always try best effort

so you want to check if any queues are depleted at which time packets can no longer be queued up for processing and will get lost

in global counters there's ctd_exceed_queue_limit for example

you can also check the soft/hardware pools :

> debug dataplane pool statistics 

and last resort you can check the packet descriptors (chip buffers)

> show running resource-monitor