Siemens PLC gives BadTooManyOperations

[Randelung]

Hey guys

I wanted to ask if you have stumbled upon this error before. I'm trying to connect to a Siemens PLC S7-1517 with a little more than 1100 nodes in (currently) one subscription. According to the FAQ it should be performant enough to handle that.
However, I'm getting BadTooManyOperations, which asyncua logs as a service fault in response to the CreateMonitoredItemsRequest.

2021-04-08 14:12:00.101 WARNING  asyncua.client.ua_client.UASocketProtocol: ServiceFault from server received  in response to CreateMonitoredItemsRequest

I'm thinking the issue may be the limited capabilities of the PLC in combination with the combined CreateMonitoredItemsRequest and PublishRequest. However, it did work when subscribing fewer than 1000 nodes, so it might be something else. From the FAQ I'm assuming that it handles subscriptions with more than 1000 nodes differently, so that may be the reason why it worked with fewer nodes... idk.

Is it possible to suppress or delay this first PublishRequest?
Any other insight?

Thanks for your input.

[AndreasHeine]

https://reference.opcfoundation.org/v104/Core/docs/Part4/5.12.2/

itemsToCreate []

how many items in the call?
have you tryed to subscribe just a chunk and in another call to add the rest?

[Randelung]

I did

        for chunk in [nodes[i:i + chunk_size] for i in range(0, len(nodes), chunk_size)]:
            results = await self.__subscription.subscribe_data_change(chunk)
            for i in range(len(results)):
                if isinstance(results[i], asyncua.ua.StatusCode):
                    self.__logger.error(f'{chunk[i].nodeid} not subscribed')
                    raise ConnectionError(f'{chunk[i].nodeid} not subscribed')

which seems to work.

[Randelung]

I think I understood my brain fart (and I'm writing this for posterity, because I'm pretty sure you know already 😄).
The call limit is not the subscription limit outlined in the FAQ. Obviously I was able to incrementally add more monitored items to the subscription and "a subscription with a maximum of 1000 elements cannot be converted into a subscription with more than 1000 elements" apparently wasn't triggered.
I thought that in order to avoid this limitation I had to subscribe to all of the nodes at once, which turned out to be untrue.
I'm also guessing that the BadTooManyMonitoredItems fault would be triggered only if the 10'000 items, or whatever is configured on the PLC, is surpassed.

Thanks again for your help.

[starturtle]

It says max items per call; mayhaps something in the server can't handle such a large request payload. Splitting sounds like a great idea. Max monitored items is the amount the server can hold, not the amount you can request in one go, afaicr.

[Randelung]

Yeah, my mistake was thinking in order to avoid the <1000 / >1000 conversion mentioned in the FAQ I'd have to subscribe all at once.

[AndreasHeine]

its best practice that the client reads some server data up front (capabilities, status, ...) and adopt to the server by only asking things which the server is able to handle otherwise you will always try catch that service faults (waist of server-ressources just for failing requests in my oppinion). a important part in opc ua is negotiation between server and client (like maxmessagsize, maxchuncksize, revised intervals, ...) but also how many nodes can i browse with one request at once, ... !