[
https://issues.apache.org/jira/browse/SLING-4009?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14161604#comment-14161604
]
Marius Petria commented on SLING-4009:
--------------------------------------
To solve 1 I can think of 2 solutions:
1. Have the agent know about the multiple endpoints by configuring multiple
importers
2. Have the agent know about multiple endpoints by configuring a list of
endpoint tags.
{code}
agent : {
"queueSplitTags": [ "one", "two", "three" ]
}
importer: {
"endpoint[one]" : "http://...4503";,
"endpoint[two]" : "http://...4504";,
"endpoint[three]" : "http://...4505";
}
{code}
The agent would add the package in multiple queues according to the split tags,
and the tag will be added as metadata to the package info.
The importer will only deliver the package to the endpoint whose tag is
mentioned in the metadata.
The reference counting I think needs to be implemented anyway.
> Multiple endpoint delivery should be fault tolerant
> ---------------------------------------------------
>
> Key: SLING-4009
> URL: https://issues.apache.org/jira/browse/SLING-4009
> Project: Sling
> Issue Type: Bug
> Components: Replication
> Reporter: Marius Petria
>
> Replication can deliver a package to multiple endpoints using a
> remoteimporter. We need a way to enable retries for a an endpoint for which
> delivery failed without delaying the delivery for the other endpoints.
> The obvious solution will be to have dedicated queues for each endpoint.
> The problems with that:
> 1. agent has no visibility over endpoints which are configured at importer
> level.
> 2. if packages can be added to multiple queues then we need a reference
> counting mechanism for packages.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
