[
https://issues.apache.org/jira/browse/SLING-5466?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15130179#comment-15130179
]
Marius Petria commented on SLING-5466:
--------------------------------------
[~jaybrown] that probably happens if the list of endpoints is changed. In order
to have a stable name for your endpoint you should give it an explicit label
{noformat}
packageImporter.endpoints = ["endpoint0=http://localhost:4503/....";,
"endpoint4=http://localhost:4504/....";]
{noformat}
> Pending distribution changes are lost if the Apache Sling Distribution Agent
> - Sync Agents Factory configuration is changed.
> -----------------------------------------------------------------------------------------------------------------------------
>
> Key: SLING-5466
> URL: https://issues.apache.org/jira/browse/SLING-5466
> Project: Sling
> Issue Type: Bug
> Components: Distribution
> Affects Versions: Content Distribution Core 0.1.4
> Environment: Windows 7 x64, AEM/CQ 6.1.0
> Reporter: Jay Brown
> Labels: patch
>
> [Using Content Distribution Core 0.1.3.r1680309 -- wasn't available in the
> list]
> When running Sling with pending distribution changes queued for a node that
> is down, a change to the Sling Distribution Agent - Sync Agents Factory
> configuration can cause the pending changes to be distributed to the wrong
> Sling node.
> It appears that changes to be distributed to a node are assigned to a generic
> "endpoint" name (not the host, itself). During the creation of the change
> nodes, the index (position) of the distribution target host in the
> packageExporter.endpoints OSGi configuration field is used to assign the
> pending changes to an endpoint named "endpoint[n])," where [n] is the
> position of the server address packageExporter.endpoints field. If
> distribution changes are queued for a host that is down AND the order of the
> packageExporter.endpoints is changed, which potentially causes endpoint[n] to
> represent a different host, the changes are then shipped off to the wrong
> host, and the original host never receives the queued changes.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
