hI yURI, On Thu, 2018-05-31 at 15:56 +0200, Yuri Simione wrote: > A customer should do an application decommissioning. We are talking > about a content management application and a repository with about > 100 million of documents/objects. > > The decommissioned app will not be updated after data > ingestion/migration and it will accessed by very few concurrent users > but just for reads operations. > > I am wondering if OAK can manage a repository with a hundred million > of nt:unstructured nodes with one binary file (few kbytes) for each > node.
That _should_ be OK ( haven't been involved in large-scale oak repositories so far ) , with the caveat that you should avoid large numbers ( >10k) of orderable child nodes below the same direct parent. Or to put it in another way - either make sure that you don't have a flat structure where one parent has > 10k children, or make sure the node type of the parent does not imply the children are orderable, e.g. use oak:Unstructured instead of nt:unstructured ( see [1] ). > > Another question: can the fulltext component be disabled during data > ingestion? Yes, just make sure that you don't create any full-text index definitions when initializing your repository. Hope this helps, Robert [1]: https://jackrabbit.apache.org/oak/docs/dos_and_donts.html#Large_number_of_direct_child_node
