Hi guys, I got back to testing UMH and this time wanted to use the umh core setup. I got it running really easy inside a docker container. But I'm having issues connecting it to my opc ua server. My tests confirmed that I can acces the ip-adres and the port from inside the container. First I had issues with creating the connection, sometimes it seems to work with my device name, sometimes with the ip-adres, but not consistently. Mostly I get: Failed to wait for protocol converter to be active: protocol converter 'l1' was removed because it did not become active within the timeout period. Please check system load or component configuration and try again But when The connection works and I want to configure the OPC UA connection I get this error: Starting_failed_dfc_missing No DFC added yet (port is open) - starting failed: no DFCs configured I tried looking in the docs and following this webinar: . But I can't find any info on these errors and how to resolve them.

hi bert! welcome 🙂 can you give us the full log that you receive when you get that error messagew? a couple of message bove that error message you get the "real" technical error

Well I deleted it and retried building the connectiona again and now I can't setup the connection neither with my device name, nor it's ip-adres. I get this in the action log: https://cdn.discordapp.com/attachments/1408055637832896708/1408060779449815152/action-log-2025-08-21T12_08_03.862Z.txt?ex=68a85e4a&is=68a70cca&hm=d7d28be964f2937b0eb943dcd01b6a8245635cb6c3cf84cb736928f8f7be1551&

The second message is helpful! This somehow looks an issue related to the volume, that we cannot access the

/umh-core-data

folder. It could be a permission issue? You could try removing the volume temporariley, which will cause umh-core to not being able to save data (or only in the container). but if it works then, then we know it has to do with the volume somehow (maybe the OS has some weird restrictyions there? here are also similar issues: https://docs.umh.app/getting-started#quick-troubleshooting

Ah yes, it does seem like an issue with the volume. I'll have to look into that. Without the volume I was able to make the initial connection. But then failed to make the opc ua connection. level=error msg="Failed to connect to opcua: error could not connect successfully to any endpoint" @service=benthos label="" path=root.input [10:46:11] - Failed during rollout: Protocol converter 'l1' edit timeout reached. It did not become active in time. Rolled back to previous configuration I probably need some additional settings based on my encription. Wanted to let you know that the link under Input is not clickable. https://cdn.discordapp.com/attachments/1408055637832896708/1408372764297134222/image.png?ex=68a980d9&is=68a82f59&hm=ebb2c41714eb1ae3fc4e13563ed2b7066ca1f4d17a4f1147b365bd420f6569be&

Good info with the link, @ricci2511 will look at it soon

This is my server: opc.tcp://DESKTOP-FKALIPH:4840 it has basic256Sha256 security with Sign & Encrypt

maybe because you have two dots in the virtual path?

Yes, but I don't get where these 2 dots come from. The path is standard from the PLC. But apparently with restarting my umh container it works.

it should be already available there, and it was developed by @runeston3

It should be in the help file on the UMH github. It is an addon i made some time ago. I use it here i @Diederik has also used it with sucess if i remember correct.

Okay, thank you. I'll ask Diederik how he did it.

I can help you as soon as i have time. Have you looked at the documentation?

Lets have a look tomorrow @bertvb1

The plugin is quite straight forward. Main thing is that you need to create a connection in the PLC to the server to allow for ads communication

Yes, the issue is not with using the plugin. But that I can’t select it inside the protocol dropdown.

oh. i see. I have not used the mgmt console so much yet. i have mostley used the old stack. However if i remember correct you should be able to manually put in the config.

When I insert something as this in the input: ads: targetIP: '192.168.3.70' # IP address of the PLC targetAMS: '5.3.69.134.1.1' # AMS net ID of the target targetPort: 48898 # Port of the target internal gateway runtimePort: 801 # Runtime port of PLC system hostAMS: '192.168.56.1.1.1' # Host AMS net ID. Usually the IP address + .1.1 hostPort: 10500 # Host port readType: 'interval' # Read type, interval or notification maxDelay: 100 # Max delay for sending notifications in ms cycleTime: 100 # Cycle time for notification handler in ms intervalTime: 1000 # Interval time for reading in ms upperCase: true # Convert symbol names to all uppercase for older PLCs logLevel: "disabled" # Log level for ADS connection symbols: # List of symbols to read from - "MAIN.MYBOOL" # variable in the main program - "MAIN.MYTRIGGER:0:10" # variable in the main program with 0ms max delay and 10ms cycleTime - "MAIN.SPEEDOS" - ".superDuperInt" # Global variable - ".someStrangeVar" This is the error I get: [10:06:57] - [left: 28 s] received log line: configuration file read error: /run/service/benthos-dataflow-read-protocolconverter-l1ads/config/benthos.yaml: (6,1) unable to infer input type from candidates: [ads]

that is strange. If i look at the uhm benthos repo the ads extenstion is imported so it should be availible.

i have the same issue running it directly from the benthos umh docker container (independent from UMH stack)

shouldnt there be a ads_plugin folder somewhere? (where it contains adsPlugin.go, go.mod and go.sum)? If you're free on this nice Saturday morning, be welcome to join our Teams meeting! 😂 https://cdn.discordapp.com/attachments/1408055637832896708/1411264227577958470/image.png?ex=68b405bb&is=68b2b43b&hm=677a57b548d78874058da7eaf284ea9843c741f0acbcb71601a97f128f66735d&

No. The plugin in in my github repo and is imported in the umh benthos build. This way umh does not need to have the responsibility for testing and maintaining the plugin. That is up to me.

this is what i do, i pull the container independantly from UMH but still the same thing:

hmm.. I am using an older container for this. It could be that the plugin needs to change some import addresses since benthos is now redpanda.

i will pull some random older versions because im pretty sure it worked before

Strange. i could build it and publish a docker container with just the plugin to see.

I'm pretty sure i tested a few months ago with 0.5.2

no. I am not sure when it was implemented in the umh.

Im searching in the discord right now 😉

in v0.3.14-beta2 the alarm processor was added and the ads was before that. so try 0.3 somthing.

doesnt seem to work either

That is really strange. I will start up my plc and build the plugin to test.

Thank you for helping out. And goodluck with the stairway!

That is really weird

i just compiled the plugin example i did. It compiles fine and benthos registed the plugin so it seems the old benthos import paths are still valid. Has there been any change in benthos on how it registers input in the latest versions @jermuk ?

@janikkn something for our todo list

What version of Benthos did you use to make it work?

The one in the repo. https://github.com/RuneRoven/benthosADS

does our analysis make sense? @jermuk @runeston3 both @bertvb1 and I tested in 2 different ways and came to the same results.

and what result do you get?

this results each time in: "configuration file read error: /benthos.yaml: (2,1) unable to infer input type from candidates: [ads]"

hmm.

Bert has the same result testing with umh core container (i obviously test with just the benthos-umh container)

there is something strange going on. The repo builds fine and when i run it, it connects fine to my PLC and reads the tags.

could you try to immitate our testing procedure?

no. benthos should reconice the input and try to connect to the plc. You should get a timeout of the connect.

Yeah for me I was able to connect to the plc. But got the error on the second step. Benthos was not in the dropdown, so I took something else and put the Ads config instead. But had the same error.

if you have the same error there is a problem with the import of the plugin in the umh release.

so you where able to reproduce our issue?

yes. I pulled the container where it first was released and i get the same problem.

SMTP does work indeed: https://cdn.discordapp.com/attachments/1408055637832896708/1411770990589120584/image.png?ex=68b5ddb1&is=68b48c31&hm=b9f55cdc422caa964e77c6739920ee73845176cb2ce43e1d5fcbe616cbb05d68& https://cdn.discordapp.com/attachments/1408055637832896708/1411770990803025992/configSMTP.yaml?ex=68b5ddb1&is=68b48c31&hm=57e5dde20a8eb67e088fb29d74e2c4c49a42ea41aa88028020f55232278ce7ea&

they a re imported in the same way.

i have been meaning to update it for a while now so it is good i finally got a push. 🙂

nice.

PR has been merged, will go out with the next benthos version release 🙂

I dont think anyone will be surprised but with 0.11.1 it works now! 😄 thanks again @runeston3 @jermuk

That's really good to hear. i will see if i can setup some automated test for it later as well and see how we can avoid problems like this in the future.. However, now there is at least a compiled and working container with it so we know that 0.11.1 will always work.

Wow guys, thank you all for the quick action on this issue. I'll make sure to test it as well on umh-core, once it's updated with the new Benthos version as well.

i am glad the plugin is useful and someone is using it. Are you running twincat 3 on the PLC?

Yes, exactly. We got some CX7000's for a small project. Last week I noticed that they can't run OPC UA, only ADS.

ok. The plugin is not yet tested with twincat 3 but it should not be any problem since the ads is the same. It is mainly the runtime ports that are different for tc3.. I have a cx7000 at home i am meaning to setup for testing.

Hi guys, I don't know what's going on, but I tried to upgrade my docker core to version v0.43.1 from v0.40.0 by changing it in my docker file. And since then I'm not able to get a connection to my plc anymore. I tried deleting everything and rebuild everything, but It's not working. Not in v0.43.1 and not in V0.40.0. Attached you can find some additional info. I don't know what else I can try to get my connection back again. https://cdn.discordapp.com/attachments/1408055637832896708/1415308757423882250/image.png?ex=68c2bc7e&is=68c16afe&hm=c68711538e8ae0267a5f13dd9e77127df28eda7cbe4591333bbf61e4ccd56fc6& https://cdn.discordapp.com/attachments/1408055637832896708/1415308758409809981/action-log-2025-09-10T12_08_41.547Z.txt?ex=68c2bc7f&is=68c16aff&hm=0538962b8a0aaf9ff2f9b42b5f19ae7614f1c3f8149ef796035a61480ace1eb9& https://cdn.discordapp.com/attachments/1408055637832896708/1415308758866984980/action-log-2025-09-10T12_01_34.379Z.txt?ex=68c2bc7f&is=68c16aff&hm=0fb011c7c3cc02012dbcba306660872ba906cd865f0a681cd2beaaa22446289a&

let me look at it

I am specifying the port as well. I'm using 192.168.1.218 or DESKTOP-FKALIPH with port 4840.

Is it the beckhoff plc?

Still the virtual one that I connect to via opc ua

"The virtual one"? Depending which tech/simulator you use you could run into problems. Many opcua servers have different ways of implementation. I remember testing with Prosys opcua sim server and a siemens S7-15xx which didnt work out of the box

I am running a VM with beckhoff twincat and am running it as a virtual PLC so I can connect to it via OPC UA. The issue is not really with OPC UA, but even the step before that it can't connect to the ip adres and port. But my colleague pointed out that it might be a networking issue that me be different from when I'm working from home or at work. So I will test some things first and then come back to you.

it sounds like you have a network problem with the VM. I have never used opcua with beckhoff so i cannot help you there, i have only used the ads for connection. What kind of vm do you use? Hyper-V, virtualbox? I have had several network problems when running hyper-v..

Hi guys, I'm sorry that I haven't gotten back to you on this yet, but I haven't been able to test it again yet at home and I'll look at it tomorrow. I'm using virtualbox.

to make the data available to the outside the best pattern here is to create another write bridge (does not exist yet) or as a workaround a standalone that takes uns as input and then exposes it to whatever systems you need

So for the standalone I need to host a kafka server myself and use it to publish from uns to it?

No, create a standalone and use uns input plugin there (it will automatically subscribe there from the UNS), and now you can choose whatever output you want and wherever you want to send it. There are even some examples in there on the top right

Does that mean there is no subscribing possible to the uns right now?

"benthos" which is called a standalone flow in the UI

So what would be in your opinion the best way to subscribe to the uns from a react application? What is the best way now and how will it be ideally in the future?

To use a standalone flow that exposes a HTTP/Websocket server 🙂

Ah, so we would have one websocket that get's all data automatically?

Exactly. Websicket/HTTP is wat more compatible with enterprise networks and enterprise computers than trying to subscribe via MQTT or similar from a react application

Why not enable direct interaction with Redpanda? Should that not be the UNS of the enterprise?

Almost all large front ends I know always use HTTP in the middle, e.g., for authentication, rate limiting, etc Also the amount of frameworks for that is much bigger

No, haven't seen it for a dedicated frontend, the websocket approach sounds good indeed for frontends

There are advantages of using benthos around the UNS, e.g, for access control or enforcing data contracts (so you can not publish everything to everywhere)

okay, interesting. Sounds like something to try.

Here you go… https://www.linkedin.com/m/pulse/uns-further-explained-matthew-parris-8pjme

@mparris I've read through your input, thanks for that. So you see Benthos as the entrypoint to the UNS. Is that as good as having a Kafka subscription? I mean can you use kafka-like functionalities like subscribing on specific topics, replay messages, guarantuee at-least-once or exactly-once delivery?

you can still acesss kafka directly, simply export port 9092 (or something) and you can access it. but we found that this is more of an antipattern because then people: 1. would ingest their own (broken) data and then wonder why there are issues everywhere. by enforcing using benthos and the UNS output we also autoamtically enforce the data contracts. in kafka, the client is responsible for ensuring the message adheres to the schema. 2. would want to connect it to a frontend, and this creates really a big mess for any compliance later. As you have then exposed the kafka port, during any audit the IT people will suddenly start asking you so many questions. What about encryption, what about role based access control, etc.. It is way easier to do this with HTTP, and then just put an NGINX in front of it with customer specific access control. 3. struggle to use Kafka (or any message broker like MQTT) properly and configure it properly. So for most use-cases it is better to use a standalone with "UNS Input" or "UNS Output" as "at-least-once" etc. is automatically built in. But you can for example not "subscribe to a single topic" in kafka. Kafka works a little bit differently in that regard. The client has to be smart and always has to poll messages. And due to the architecture of Kafka, we technically only have a single kafka topic and work with different kafka keys. (MQTT is the opposite in that regard, the clients are dumb but then you also cannot use any replay or anything like that). With benthos and our plugins, we abstracted all of that complexity away.

Oh okay, so the internal kafka is actually before the the data of the UNS, so that's why the topics of _historian. ar enot found?

What do you mean?

I'm not familiar with Kafka and what they offer with their interface. I susepect that it will not be able to serve the needs of an "Information Consumer", which will query the platform: "list the instances of Object Type ABC, which represents a machine" Now, let me subscribe to each of those objects. I suspect that Kafka (llike MQTT) is too raw of an interface, only concerned with allowing access to the data it has in a raw format, and not in a contextualized format. Does Kafka allow you to work with Information Models, and translate the knowledge graph that would be created within the models?

@jermuk thanks for that context! It helps to understand 😁

Would a solution be to add an env var while deploying the container to enable port 9092? this way powerusers can access the kafka while it can be closed to enforce datacontracts?

No need for environment variables 🙂 just expose the port 9092 in the docker run command

Can we actually already do this UNS-discovery through the websocket? Are there any docs on that?

Ok cool. So if we want to discover the UNS, we can hit that graphql api, if we want realtime data in our frontend app, we can use Benthos with a websocket, correct?

yes! we are even thinking of providing a central API around it, but in its core that API would also just trigger the other ones

Ok, I think I understand your thought proces. One more thing. What is the recommended way to write back to the UNS? A seperate benthos flow with UNS output I guess?

I’ve pushed I3X to have a central API to discover, as well as read, write, AND subscribe. I’ll continue the conversation in the other thread about I3X

yes! and then coming from a HTTP server again, process the message and then put it into the UNS.

@runeston3 When you run it yourself with the ads protocol. Are you running it in a container? If so, can you give me some info of your configuration? What it's your container settings, what did you put in your PLC as route, what does your ads config look like? I think the plugin is working, but I now have more of a networking issue between the container and the PLC.

You need to create a route in the PLC to the server for it to work. The ads is a bit odd in that case.

I finally got some data through. But I only got some initial values when I check the history I got about 10 logs with values and then it started logging empty strings for some reason. This is my config: materiaalvoorziening: connection: nmap: target: '{{ .IP }}' port: '{{ .PORT }}' dataflowcomponent_read: benthos: input: ads: hostAMS: 192.168.1.62.1.1 intervalTime: 1000 readType: interval runtimePort: 851 symbols: - MAIN.ToggleBit - GVL_Signals.AI_VacBak_Vacuum - GVL_Signals.DI_Alarm_SawMachine - GVL_Alarms.Alarm_VacBakVacuum_High - GVL_Alarms.Alarm_VacBakVacuum_Low targetAMS: 192.168.1.177.1.1 targetIP: 192.168.1.177 targetPort: 48898 upperCase: false pipeline: processors: - tag_processor: defaults: | // Simple ADS data processor msg.meta.location_path = "{{ .location_path }}"; msg.meta.data_contract = "_historian"; msg.meta.tag_name = msg.meta.symbol_name; msg.payload = msg.payload; //does not modify the payload return msg; buffer: none: {}

hmm. interesting. try output it to console with a simple processor like {root = this} instead and check what you get to see if the input is working correct.