You are viewing a plain text version of this content. The canonical link for it is here.
Posted to oak-issues@jackrabbit.apache.org by "Manfred Baedke (Jira)" <ji...@apache.org> on 2024/02/09 13:52:00 UTC
[jira] [Commented] (OAK-10642) Add tests for operations on very large ordered collections
[ https://issues.apache.org/jira/browse/OAK-10642?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17816072#comment-17816072 ]
Manfred Baedke commented on OAK-10642:
--------------------------------------
Started with
{code:java}
@Test
public void moveOrderableWithManyChildren() throws Exception {
int max = 10000;
Session session = getAdminSession();
Node test = session.getRootNode().addNode("test-0", "nt:unstructured");
session.save();
for (int k = 0; k < max; k++) {
Node node = test.addNode(UUID.randomUUID().toString(), "nt:unstructured");
test.orderBefore(node.getName(), null);
session.move("/test-" + k, "/test-" + (k + 1));
}
session.save();
}
{code}
hoping to trigger branch commits and eventually an exception from Mongo due to exceeding the max doc size. I set update.limit=100 and tested with increasing numbers of children. Performance drops quickly, with 10000 children the test is now running for 20h. No signs of other problem though. I'm debugging to see if and how branch commits might or might not be triggered,
> Add tests for operations on very large ordered collections
> ----------------------------------------------------------
>
> Key: OAK-10642
> URL: https://issues.apache.org/jira/browse/OAK-10642
> Project: Jackrabbit Oak
> Issue Type: Test
> Components: core, jcr
> Reporter: Manfred Baedke
> Assignee: Manfred Baedke
> Priority: Major
>
> The ordering is persisted in a single string property containing the names of the children in the correct order, which is prone to run into property size limits or even document size limits and also may perform poorly.
> One concern in particular is the possibility to run into Mongo's document size limit of 16M with long transactions containing many operations that change the ordering, eventually triggering branch commits that will be stored in a single Mongo document quickly growing in size because of the weird data model used to persist the ordering.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)