You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@nifi.apache.org by "Matt Gilman (JIRA)" <ji...@apache.org> on 2015/02/19 16:58:12 UTC

[jira] [Created] (NIFI-367) Node Reconnection Issue

Matt Gilman created NIFI-367:
--------------------------------

             Summary: Node Reconnection Issue
                 Key: NIFI-367
                 URL: https://issues.apache.org/jira/browse/NIFI-367
             Project: Apache NiFi
          Issue Type: Bug
          Components: Core Framework
            Reporter: Matt Gilman
             Fix For: 0.1.0


Currently, when a Node connects to a cluster it verifies it's current flow with that of the cluster. If there are discrepancies in the flow fingerprint the connection request fails. What the Node uses as it's current flow depends on it's state. A Node that is not initialized will utilize the flow.xml on disk. A Node that is already initialized will use the current state of it's FlowController.

The issue here is that when an initialized Node connects to the cluster the flow fingerprint will differ from the cluster's whenever new properties have been added to a component (upgrading from one version to another) in their flow. The issue would exist until the cluster flow has been saved which is triggered by a modification to a flow. Initialized Node's will reconnect via user action or reconnection requests (typically caused by lack of heartbeat).

The issue here will occur after upgrades when a NiFi instance is clustered, components used in the flow have new properties, and the initialized Node connects to the cluster (typically by lack of heartbeat or user action). This won't occur during the restart after the upgrade because the Node will use the flow.xml on disk.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)