[ https://issues.apache.org/jira/browse/SLING-9482?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17119097#comment-17119097 ]
Timothee Maret commented on SLING-9482: --------------------------------------- Done in [c7814657bb42996558a781d749a93c72437c89fc|https://github.com/apache/sling-org-apache-sling-distribution-journal/commit/c7814657bb42996558a781d749a93c72437c89fc]. The change introduce the following repository operations: * One read op when a publisher agent starts * At most one write operation per cluster every 15 minutes, 0 write op if no package was produced during the last 15 minutes > Seed the cache from offset persisted in the source repository > ------------------------------------------------------------- > > Key: SLING-9482 > URL: https://issues.apache.org/jira/browse/SLING-9482 > Project: Sling > Issue Type: Bug > Components: Content Distribution > Reporter: Timothee Maret > Assignee: Timothee Maret > Priority: Major > > The approach taken in SLING-9460 to avoid sending seeding messages does not > address some scenarios like publishing content without subscriber agents. > Without subscriber agents. > To be sure we always have a recent seed available, we should persist seed > offsets in the source repository (typ. author) and seed caches from it. We do > have the local store class that allows to easily write the offsets in the > repository. To not stress the repository too much, we should batch those > writes (e.g. 1 offset update every 10 package processed). To support a > cluster, the writes must be initiated only from the cluster leader instance. -- This message was sent by Atlassian Jira (v8.3.4#803005)