Automatic scaling of Y-axis of `nats server graph` sometimes works incorrectly?

[alexbozhenko]

Observed behavior

Not sure if sudden change of values triggers it, or what:

Expected behavior

Should not show negative values and EiB on the Y-Axis?

Server and client version

# nats --version
v0.2.5-0.20250730153159-37671e28664b

# nats-server  --version
nats-server: v2.11.6

Host environment

No response

Steps to reproduce

Run:

nats server graph

In a separate terminal, run nats bench kv put, and observe the graphs for several seconds.

Not reproducible 100% of the time, but it seems like timing between running the two commands may play a role...

[alexbozhenko]

So I traced it. When nats server graph is executed without optional server argument, we do SYS.REQ.SERVER.PING.VARZ.

So we will/may collect info from a different server on every tick.
For example, here is how values look like for number of bytes from different servers:

[447944896190,"s1"]
[448273773782,"s1"]
[9534631,"s2"]
[448952430425,"s1"]
[449259224306,"s1"]
[449561064247,"s1"]
[9544549,"s2"]
[450184161945,"s1"]
[450489159237,"s1"]
[450785299594,"s1"]
[451079850207,"s1"]

And so when we compare previous and current reading from different servers in

natscli/cli/util.go

Line 929 in 5c1d0ef

func calculateRate(new, last float64, since time.Duration) float64 {

we will big negative rate.

It seems like server name/id should be a mandatory argument for this command, otherwise these graphs displaying values from different servers wouldn't make sense?

[ripienaar]

Yeah that make sense, should be required