Multiple listeners if many can ids

[erwinschrodinger1]

Description

I am building an application that needs to handle multiple ISO-TP request/response pairs simultaneously (multiple (txid, rxid) address pairs).

To achieve this, I am currently creating one NotifierBasedCanStack per address pair and attaching all of them to a single CAN bus and notifier.

---

Current Implementation

self.sockets: Dict[str, isotp.NotifierBasedCanStack] = {}

self.bus = can.Bus(interface=can_interface, channel=channel, bitrate=bitrate)
self.notifier = can.Notifier(self.bus, [])

for name, (txid, rxid) in addresses.items():
    sock = isotp.NotifierBasedCanStack(
        bus=self.bus,
        notifier=self.notifier,
        address=isotp.Address(
            addressing_mode=isotp.AddressingMode.Normal_29bits,
            txid=txid,
            rxid=rxid,
        ),
    )

    sock.start()

---

Problem

When scaling beyond ~10 (txid, rxid) pairs, I am also running a WebSocket server alongside the ISO-TP handling.

Under this load:

1. The WebSocket connection becomes unstable
2. It occasionally disconnects unexpectedly during operation

This leads me to suspect that the issue is not limited to ISO-TP performance alone, but that the overall process (or event handling/thread scheduling) is being overloaded or blocked.

---

Questions

1. Is this usage pattern (multiple NotifierBasedCanStack instances sharing a single notifier) aligned with the intended design of the library?

2. Is there a recommended way to handle multiple ISO-TP sessions concurrently on the same CAN bus?

---

Expected Guidance

I’m trying to understand the intended scaling model of the library and the recommended architecture for handling multiple concurrent ISO-TP communications efficiently.

[pylessard]

Hello

1. It is the intended deisgn
2. You do it like it should I think

If you have performance issues, I suggest to run a profiler. I strongly suggest pyinstrument, make a html report and look at the flame graph. You'll see where the bottleneck is.

Please, report the outcome here.

pip install pyinstrument
python -m pyinstrument -r html -o report.html myfile.py

Open report.html with a browser and select the "timeline2 radio box at the top.