I tried to attach images to the body of the email but they are in the attached 
files down below, sorry for the inconvenience.

On 2025/11/10 11:18:08 Paminos Sym wrote:
> Hello everyone,
>
> Let me start by describing how we use NiFi: We have many different types of 
> flows of which exist multiple instances, each with its own parameter context. 
> For example, flow A may have 15 instances and flow B another 15. Each flow is 
> version controlled using NiFi Registry and is composed of a few process 
> groups that are usually each also version controlled. Flows go into a flow 
> bucket, while the process groups they are made up of go into a process group 
> bucket. Essentially it is similar to having an Abstract object (flow) 
> composed of other abstract objects (internal process groups). Every time we 
> create an instance, this object and the ones composing it are all 
> instantiated by getting parameters.
>
> [cid:1526dc04-96ea-4126-9b7d-e1f9bef73ad1]
>
> Until our recent upgrade to 2.6.0 this was functional and working mostly as 
> expected. To create a new instance of Flow A I would get one from the 
> registry, change the name, create a new parameter context, and I very quickly 
> had a new clone of flow A with different parameters. Now when I get a new 
> instance from the registry with "Keep existing Parameter Contexts" checked I 
> get this error:
> [cid:96253bad-a86f-4df6-b0ae-d393226be779]
> (I can't share the parameter context names)
> I believe this has something to do with how our parameter contexts are 
> organized. There is a "Root Parameters" context which both Flow A and Flow B 
> inherit. Flow A and Flow B both have their "super" contexts (like 
> "super-flow-a-parameters") which inherit the Root Parameters context, and the 
> super contexts are inherited by the instance specific context. Then each 
> instance and the process groups within will use the instance specific 
> context. This base.sourceType parameter belongs to an instance specific 
> parameter context and is inherited from the root parameters context, and is 
> blocking my operation, but the issue is that there is no reason for this 
> parameter to be updated when I create a group. Creating a new group shouldn't 
> change the value of a parameter, so why is it trying to update this one?
>
> To bypass this I create the group with the "Keep existing Parameter Contexts" 
> unchecked, however this creates copies of some contexts (e.g. 
> "flow-a-instanceA (1)", "flow-a-instanceA (2)").
>
> [cid:3a645302-57ad-4138-99c0-5a12a054dbe5]
>
> After deleting the superfluous copies of the parameter contexts and creating 
> the new instance's parameters I try to change the whole group's parameters. 
> Reminder: This is the instance's group, which is under Flow A's version 
> control, and contains process groups that are also version controlled. 
> Before, to do this I would simply click "configure" on my group, select my 
> parameter context, select "Apply Recursively", and after pressing apply it 
> would work as expected. Now I get an error:
> [cid:9e6f1b00-6421-419e-bffa-e03b5f6db4e7]
> This seems very strange, not only because this used to work without any 
> issue, but also because switching parameter contexts on a versioned flow 
> being a problem seems completely wrong. My work-around for the time being is 
> to uncheck "Apply Recursively" and manually apply the correct context in each 
> of the internal process groups, which is time consuming and error prone.
>
> The final issue I have encountered thus far is more minor. Previously, if I 
> made a change to a versioned flow and wanted to update every flow with that 
> version I would simply go to each one, running or not, and select "change 
> version". Though time consuming it was a frictionless procedure. Now I get an 
> error similar to the first one, and the work-around is to stop the flow I'm 
> updating, change flow version, and start it again. This makes it even more 
> time consuming but at least it's more manageable.
>
> I find it hard to believe that this is intended behaviour. Is it because we 
> updated from 2.2.0 straight to 2.6.0? Is this something others have 
> experienced?
>
> Any help or explanation would be greatly appreciated, thank you.
>
> Epameinondas Symeonidis,
> Ianic S.A.
>
>

Reply via email to