You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@nifi.apache.org by "Andre F de Miranda (JIRA)" <ji...@apache.org> on 2017/03/27 01:56:41 UTC

[jira] [Commented] (NIFI-87) Configuration across nodes in a cluster should be identical including properties files and more

    [ https://issues.apache.org/jira/browse/NIFI-87?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15942525#comment-15942525 ] 

Andre F de Miranda commented on NIFI-87:
----------------------------------------

[~joewitt] - I appreciate the problem but I must say the removal of those properties will cause little comfort to those using scalable clusters as the following underlying non-scalable issues remain:

- secure clusters must have certificates issued to each of the nodes, this is also true to ephemeral / auto-scaling nodes
- same thing applies to kerberos principals (if used for authentication)

Furthermore, given that most cluster nodes would be using some sort of configuration management tool, the automatic setup of the three above properties is trivial to things like Puppet and co.

> Configuration across nodes in a cluster should be identical including properties files and more
> -----------------------------------------------------------------------------------------------
>
>                 Key: NIFI-87
>                 URL: https://issues.apache.org/jira/browse/NIFI-87
>             Project: Apache NiFi
>          Issue Type: Improvement
>          Components: Core Framework
>            Reporter: Joseph Witt
>
> Currently each node must have different values in its nifi.properties file for the hostname
> nifi.web.http.host
> nifi.web.https.host
> nifi.cluster.node.address
> In order to be scalable it must be the case that all nodes can have the exact same configuration.  This should be achieved by removing any reference to the hostname from the properties file and instead use the name of the Network Interface (or interfaces)
> nifi.web.http.interfaces=etho0,eth1,...
> Generally, in order to quickly scale up and down VMs will be used and it is considered acceptable to expect that all VM's should use the same NIC for comms.  This means we must fully understand exactly how these properties are used.  They are used in the JettyServer class as well as in clustering.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)