Standalone flow doesn’t reconnect automatically after umh-core restart United Manufacturing Hub #umh-support

Standalone flow doesn’t reconnect automatically af...

zestones

11/04/2025, 8:42 AM

Hello, I’m currently using a standalone flow to store topics in my historian database (

kafka_to_postgresql_historian_bridge

). However, I've noticed that when my

umh-core

instance goes down and I restart it, the standalone flow doesn’t reconnect automatically. The only way I’ve found to re-establish the connection is to manually update the configuration (for example, in the input.yaml I change the local IP in

seed_brokers

to localhost or 127.0.0.1) and then click Save & Deploy. Otherwise, the flow remains inactive. Is there a way to make it reconnect automatically once the

umh-core

instance is back up ? In the logs i only see one attempt at connecting :

Jermuk

11/04/2025, 10:42 AM

in the logs you can actually see that it initially failed during restart (probably because redpanda wasnt started up yet), but then it connected (see last log line)

Jermuk

11/04/2025, 10:42 AM

or what do you mean?

zestones

11/04/2025, 12:33 PM

Yes, but it doesn't publish data to the database after a restart unless I update the config file. For example, the data here stopped at 12:19:15, and even after restarting the UMH instance and waiting a long time, it did not start publishing the data again.

Jermuk

11/04/2025, 12:52 PM

@krgrsebastian did you also experience this issue? Might be also important for our customers?

Jermuk

11/05/2025, 12:16 PM

we are currently looking at this as the full dev team and here are our notes: 1. you hsould ideally use the uns input, and not kafka_franz 2. look for same consumer group and make them unique. otherwise they will share the same messages, and maybe there is another floiw in parallel with the same consumer group 3. we dont have this happening on our siode, but could also be a weird edge case bug 4. there is also some inconsistency in the UI. "Starting" state actually means "Confirming Startup", because it can already be started but we couldnt verify it that iwas successful because there were errors/warnings during startup maybe that helps, esepcially point 2?

Jermuk

11/05/2025, 12:17 PM

also we will work on the ability to have this automatied as almost all users of UMH will need to write data to timescale, @daniel_umh will work in the next weeks/months on it

Jermuk

11/05/2025, 12:17 PM

greetings from the entire dev team here 🙂

zestones

11/05/2025, 1:28 PM

thank you for your support ! I tried your suggestions and it worked ! - my consumer group is unique, so no conflicts with other flows - I switched the input from kafka_franz to uns like this:

Copy code

yaml
input:
  uns:
    broker_address: localhost:9092
    consumer_group: kafka_to_postgresql_historian_bridge
    umh_topic: umh\.v1\..*_historian.*

now, when I stop and restart umh-core, the standalone flow reconnects automatically and publishes data to timescale without any issues.

zestones

11/05/2025, 1:31 PM

so yeah I dont know why it didn't work for the kafka_franz maybe a weird edge case bug as you said..

zestones

11/05/2025, 1:33 PM

Great ! Many thanks to you and the whole dev team, looking forward to this improvement 🙂

daniel_umh

11/05/2025, 2:46 PM

awesome to hear it works 😄

3 Views

Previous Next