When the breaking changes will be stopped in Zeebe

Hi Team,

I want to know from when the breaking changes will be stopped in every release? Because I dont want to work with my team for 2 -3 weeks to do migration every time, its lot of work and problem for maintenance.

Please let me know from which version breaking changes will be stopped else
unfortunately I have to switch the framework from zeebe to some other, because I am running zeebe in production from past 6 months

1 Like

Hi @regojoyson,

I’m very sorry for the inconvenience. We are aware of this problem and try our best to avoid it.

With 0.23.0, we introduced new mechanics in the broker to be backward-compatible and to allow rolling updates.

Of course, Zeebe is still in an early stage and we can not guarantee that it will never happen again (if there are good reasons for doing it). But we will definitely try to avoid it - or make the migration as smooth as possible.

So, stay tuned :wink:

Best regards,
Philipp

3 Likes

Hi @regojoyson,

For Zeebe 0.23.0 in particular, we provide a migration tool to automate the migration of the BPMN models from JSONPath to FEEL.

The change of the configuration format should be fairly transparent to you if you use the Helm chart for deploying Zeebe.

Cheers,
Falko

2 Likes

@falko

Ya I Agree but my zeebe having nearly 400 instances runing which will need 1 month to end and also daily min 4 to 5 instances is created.

And also we are using spring Kafka for messaging with zeebe , so it’s very difficult to migrate.

If we have to create new zeebe instance, we have to do kafka also and another pain is we have to manage both for 1 month untill all instances are completed .

1 Like

Thanks for the feedback on your use case @regojoyson - we are discussing internally how we can reduce breaking changes and mitigate the ones that do come in.

1 Like

Hi @jwulf this is also interesting for us - for zeebe to be used in production there has to be a simple way of upgrading the broker without harming running instances (workflows are less of a problem than workflow instances in that regard).
P.S. Most cluster products offer zero-downtime upgrading in-place node-by-node

1 Like