Understanding FreeRADIUS statistics

[Alan DeKok]

On Jun 23, 2024, at 10:59 AM, dominic.stalder at unibe.ch wrote:
> As already written, I updated to FreeRADIUS 3.2.5 on Friday evening and was now able to do some tests in our lab environment. I think there are still some "glitches" and I still have some comprehension questions.

  OK.

  The main reason for the counters to be imperfect is that it takes time for the server to process packets.  So if you query a running server, you might get more "requests" than "responses".  But the numbers should be close.

  Also, if the server stops receiving packets, and cleans up its internal de-duplication cache, the number of requests should match the number of responses.

  That being said, some simple tests show that the stats are off.  :(  I've pushed some patches which:

a) don't count Status-Server as Access-Request

b) in some cases other packets were double-counted as Access-Request, so those numbers were too high

c) it now updates the response statistics more quickly if it can.

  So the numbers should be a lot better.

> Thanks in advance for any help on this topic; I know you wrote that the statistics are not always perfect, but I would like to build a monitoring on some parts of those values and if they could be corrected, that would be cool.

  Yup.

  Alan DeKok.