BadSessionNotActivated issue

[m4rkisgrov3]

Hello, please can anyone help?
I'm using the python-opcua library in a fairly simple way, so I've been surprised when I encountered this error.

I've written a Python client that subscribes to about 160 nodes. I create 1 handler, and subscribe to all the nodes in one go. As follows:

    self.client = Client(self.opcserverurl)
    self.client.connect()
    validnodes = []
    ...
    _In a for loop I now populate the valid nodes list:_
    for...
            node = self.client.get_node(nodename)
            validnodes.append(node)
    ...
    handler = SubscribeHandler()
    sub = self.client.create_subscription(100, handler)
    sub.subscribe_data_change(validnodes)

Of course, when any of the node values change, the handler is called. This has worked quite happily until I ramp up the number of nodes I subscribe to; I now subscribe to 160, I thought I'd be able to subscribe to 1000s. Also, note the 100 period value in create_subscription.

So, now when I run this whilst subscribing to more nodes, it falls over on the subscription with this error message:

_exception calling callback for
Traceback (most recent call last):
File "/usr/local/lib/python3.8/concurrent/futures/_base.py", line 328, in _invoke_callbacks
callback (self)
File "/root/.local/lib/python3.8/site-packages/opcua/client/ua_client.py", line 493, in call publish callback
self._uasocket uasocket.check_answer (data, "while waiting for publish response")
File "/root/.local/lib/python3.5/site-packages/opcua/client/ua_client.py", line 93, in check_answer
hdr. ServiceResult.check()
File "/root/.local/lib/python3.5/site-packages/opcua/ua/uatypes.py", line 218, in check
raise UaStatusCodeError (self.value)
opcua.ua.uaerrors.auto.BadSessionNotActivated: "The session cannot be used because Activate Session has not been called." (BadsessionNotActivated)

Note the error in the traceback, BadSessionNotActivated.

Please can someone help, with less nodes I subscribe to, all this works well and all data changes are captured, when I subscribe to more than 160 nodes, I get this error.

I'm thinking this could be a timing issue? I specify a value of 100 for the period when I create the subscription, is this enough? Or, another thing that crossed my mind is that should I be creating a session since it's telling me BadSessionNotActivated?

Many thanks for help, happy to provide more info if required.

[AndreasHeine]

what is the opcua server?
how many sessions does the server allow?
did you check the servers capabilities?
how many monitored items does he allow per subscription?
have you checked wireshark trace to get a clue who sends the error (i assume the server)?

[m4rkisgrov3]

So, is this a server side issue or client side?
My client is calling a callback? The handler is this? But it cannot call it? Is that correct?
Thanks.

[AndreasHeine]

the subhandler callback is called for every datachangenotification from server!
dont know whats inside your callback...

just for testing change the Subhandler:

class SubHandler(object):

    """
    Subscription Handler. To receive events from server for a subscription
    data_change and event methods are called directly from receiving thread.
    Do not do expensive, slow or network operation there. Create another 
    thread if you need to do such a thing
    """

    def datachange_notification(self, node, val, data):
        print("Python: New data change event", node, val)

[m4rkisgrov3]

Yeap, I'll try this. Good point.
I do have quite a bit going on in the callback... I create a new thread and call a REST API endpoint within it, perhaps I'm doing too much in here?
Appreciate the help here.

[swamper123]

Do not do expensive, slow or network operation there. Create another 
thread if you need to do such a thing

With "to much" it is meant, that you shouldn't run long going processes (someone once wanted to make imageprocessing, which is slowing down the server really fast).

OSError: [Errno 9] Bad file descriptor

Sounds more like another problem with your socket ... are you handling that correctly?

[m4rkisgrov3]

Thanks everyone for help here.
I've suppressed the problem and narrowed down the issue so thought it was worth noting this in the forum.

So when I'm subscribing to lots of nodes (circa 500), when I start my OPC client I get an initial surge of data change notifications. This would mean I get 500 calls to my callback at pretty much the same time. This isn't good and causes it to break with the errors above. So, to suppress this issue, I've put in a 10 seconds delay into my subscribe handler meaning that it doesn't process any nodes during this initial surge (or in the initial 10 seconds of starting the OPC client). Not ideal and it feels like a fudge, but it means I don't get that massive ramp up in activity to deal with. After the delay, it ticks along nicely and processes everything that comes in.
I'd prefer a more robust fix of which I'm still looking to explore...

@oroulet are these issues you've encountered before?

I appreciate everyone's help on this. :)