nifi-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Mark Payne (JIRA)" <>
Subject [jira] [Commented] (NIFI-1743) Allow Heartbeats to be replicated to all nodes in the cluster
Date Wed, 18 May 2016 17:30:13 GMT


Mark Payne commented on NIFI-1743:

[] - sorry I just saw your comment. Yes, it is part of the Zero
Master Clustering. The idea is that ZooKeeper will be used to elect a 'cluster coordinator'
and publish who this is so that nodes are able to push heartbeats to the currently elected
coordinator. This work was done actually and pushed to master but was done under NIFI-1678.
I marked this ticket as resolved, since it was merged to master and added a "Related to" link
to NIFI-1678.


> Allow Heartbeats to be replicated to all nodes in the cluster
> -------------------------------------------------------------
>                 Key: NIFI-1743
>                 URL:
>             Project: Apache NiFi
>          Issue Type: Task
>          Components: Core Framework
>            Reporter: Mark Payne
>            Assignee: Mark Payne
>            Priority: Blocker
>             Fix For: 1.0.0
> Originally, we had planned to use ZooKeeper to store node heartbeats. However, this caused
quite a bit of concern within the community and provided a fairly poor user experience if
we lost the quorum, as the state of the entire NiFi cluster is no longer known since we are
unable to read from ZooKeeper.
> As an alternative, we will instead send a heartbeat message to each node in the cluster.
The Elected Leader will be responsible for determining if a particular node is to be disconnected
due to a lack of heartbeat. All other nodes will then be notified of this; if Node X is deemed
disconnected and then sends a heartbeat to Node Z, Node Z should respond by notifying Node
X that it was disconnected. If Node X then receives many of these responses because it is
heartbeating to many nodes, that is okay - it should be implied idempotent behavior.

This message was sent by Atlassian JIRA

View raw message