[
https://issues.apache.org/jira/browse/CONNECTORS-921?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14146945#comment-14146945
]
Karl Wright commented on CONNECTORS-921:
----------------------------------------
My thought on how best to handle this:
(1) Assume that the most likely situation is true (e.g. the root path exists)
(2) If a NoNode zookeeper exception is thrown, then create the node, and if
successful or "node already exists", then proceed
This could effectively limit the performance impact, at the expense of code
complexity. In cases where a permanent child node of the root node needs to
be created, then there's an ambiguity when NoNode is thrown; because there's no
way to tell if it is the root node or the child node that is missing. So in
those cases, the back-off is more challenging, with multiple node creations
attempted in specific orders.
> Refine manifoldcf path in Zookeeper
> -----------------------------------
>
> Key: CONNECTORS-921
> URL: https://issues.apache.org/jira/browse/CONNECTORS-921
> Project: ManifoldCF
> Issue Type: Improvement
> Components: Framework core
> Affects Versions: ManifoldCF 1.6
> Reporter: Graeme Seaton
> Assignee: Karl Wright
> Priority: Minor
> Fix For: ManifoldCF 2.0
>
>
> The nodes created in Zookeeper by MCF are created directly beneath the root.
> This can make it difficult to find the nodes for other applications. There
> seems to be a convention to create an application node under the root (based
> on Hbase and Hadoop usage).
> I would like to create a manifoldcf 'parent' and place all of the nodes
> within it. It might also make sense to create 'parent' nodes for each of the
> PATH_PREFIXES.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
