[
https://issues.apache.org/jira/browse/NIFI-190?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16169624#comment-16169624
]
Minmin DU commented on NIFI-190:
--------------------------------
Hi there,
I am recently working in a project that uses NiFi a lot. Based on my
understanding, the current Notify / Wait (in NiFi 1.3) is working in a way that
relies on count. It provides capacity to solve 1:N dependency whereas I don't
think it works well for M:N dependency.
For example, FlowFile D depends on FlowFile A and FlowFile B. And FlowFile E
depends on FlowFile B and FlowFile C. It means that FlowFile D won't start
until both FlowFile A and FlowFile B come in. After FlowFile D kicks off,
FlowFile B cannot be removed from distribute cache map since FlowFile E also
relies on it.
In our project, we had to create custom Wait processor to handle this multiple
dependency. I guess multiple dependency scenario happens a lot in an enterprise
workflow scheduling.
Could you please let me know if the existing Wait/Notify processors can deal
with this scenario?
Thanks a lot.
Cheers,
> Wait/Notify processors
> ----------------------
>
> Key: NIFI-190
> URL: https://issues.apache.org/jira/browse/NIFI-190
> Project: Apache NiFi
> Issue Type: New Feature
> Components: Extensions
> Reporter: Joseph Gresock
> Assignee: Joseph Gresock
> Priority: Minor
> Fix For: 1.2.0
>
> Attachments: Wait_Notify_template.xml
>
>
> Our team has developed a set of processors for the following use case:
> * Format A needs to be sent to Endpoint A
> * Format B needs to be sent to Endpoint B, but should not proceed until A has
> reached Endpoint A. We most commonly have this restriction when Endpoint B
> requires some output of Endpoint A.
> The proposed Wait/Notify processors enable this functionality:
> * Wait: routes files to the 'wait' relationship until a matching Release
> Signal Identifier is found in the distributed map cache. Then routes them to
> 'success' (unless they have expired)
> * Notify: stores a Release Signal Identifier in the distributed map cache,
> optionally with attributes to copy to the outgoing matching Wait flow files.
> An example:
> Wait is configured with Release Signal Attribute = "$\{myId}". Its 'wait'
> relationship routes back onto itself.
> flowFile 1 \{ myId : "123" }
> comes into Wait processor
> Wait checks the distributed cache map for "123", doesn't find it, and is
> routed to the 'wait' relationship
> Notify is configured with Release Signal Attribute = "$\{myId}"
> flowFile 2 \{ myId : "123" }
> comes in to Notify processor
> Notify puts an entry in the map for "123" with any other attributes from
> flowFile2
> Next time flowFile 1 is processed by Wait...
> Finds an entry for "123"
> Removes that entry from the map
> Copies attributes to flowFile 1
> Sends flowFile 1 out the success relationship
> Notify will optionally cache attributes in the distributed map, as determined
> by a regex property. This is what allows the output of Endpoint A to pass to
> Endpoint B, above. Wait also allows conflicting attributes from the cache to
> either be replaced or kept, depending on property configuration.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
