Description

Customer experienced overall latency with their connections from AWS into vSRX and after considerable digging, found the switch port the vSRX is connected to got the "output discards" increment.

 

#sh int port-ch 20 | inc discard

0 input errors, 0 input discards

0 output errors, 3380987 output discards

 

#sh int port-ch 20 | inc discard

0 input errors, 0 input discards

0 output errors, 3381140 output discards

 

The vSRX was recently upgraded and in attempt to make things better, but it has not, the discards continue to increment with traffic.

Solution

Having reviewed all the logs collected from the vSRX, we don't see any software or configuration on vSRX side which could cause the discard on switch side.

1.The "output discards" increment definitely means that the switch port experiences congestion, when the forwarding class queues continue to receive their full buffer allocation, it will start to discard the packets.

2.By default, four queues are assigned to four forwarding classes, we may check the drop counter for each queue, for example, below is the output on Juniper QFX switch.

 

root> show interfaces extensive no-forwarding

 

 Output errors:

  Carrier transitions: 0, Errors: 0, Drops: 325366, Collisions: 0, Aged packets: 0, FIFO errors: 0, HS link CRC errors: 0, MTU errors: 0, Resource errors: 0

 Egress queues: 8 supported, 4 in use

 Queue counters:    Queued packets Transmitted packets   Dropped packets

  0           88142147228     88142147228          0

  1            9454572139      9454247120        325019

  2            438723733      438723386         347

  3             3713348       3713348          0

 

From the output, drops seen are only for Queue 1 and 2: 325019 + 347 = 325366, the amount of drops for this interface can be determined due to the lack of Class-of-Service configuration or schedulers configuration needs to be adjusted( add more percentage to Queue 1 and 2).

 

3.If all the 4 queues are dropping the packets, then we may try to increase the buffer size of the IOC card, for example, on Juniper QFX switch, the config is like below.

 

set chassis fpc 0 pic 0 q-pic-large-buffer

 

4.If drops are still seeing after implementing the steps above, then it's good to upgrade to 10G network ports for this vSRX deployment. 

Modification History

2024-06-07 : Article Created