You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@bookkeeper.apache.org by "Sijie Guo (JIRA)" <ji...@apache.org> on 2017/06/12 15:22:00 UTC
[jira] [Resolved] (BOOKKEEPER-1096) When ledger is deleted, along
with leaf node all the eligible branch nodes also should be deleted in
ZooKeeper.
[ https://issues.apache.org/jira/browse/BOOKKEEPER-1096?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Sijie Guo resolved BOOKKEEPER-1096.
-----------------------------------
Resolution: Fixed
Fix Version/s: 4.5.0
Issue resolved by merging pull request 186
[https://github.com/apache/bookkeeper/pull/186]
{noformat}
commit 07852d35856dca232450135913090bac27b29abe
Author: Charan Reddy Guttapalem <cg...@salesforce.com>
AuthorDate: Mon Jun 12 08:20:53 2017 -0700
Commit: Sijie Guo <si...@apache.org>
CommitDate: Mon Jun 12 08:20:53 2017 -0700
BOOKKEEPER-1096: recursive znode delete
When ledger is deleted, along with leaf node
all the eligible branch nodes should be
deleted in ZooKeeper.
Author: Charan Reddy Guttapalem <cg...@salesforce.com>
Reviewers: Sijie Guo <si...@apache.org>
Closes #186 from reddycharan/recursiveznodedelete
{noformat}
> When ledger is deleted, along with leaf node all the eligible branch nodes also should be deleted in ZooKeeper.
> ---------------------------------------------------------------------------------------------------------------
>
> Key: BOOKKEEPER-1096
> URL: https://issues.apache.org/jira/browse/BOOKKEEPER-1096
> Project: Bookkeeper
> Issue Type: Improvement
> Reporter: Charan Reddy Guttapalem
> Assignee: Charan Reddy Guttapalem
> Fix For: 4.5.0
>
>
> Currently when we delete a ledger, we delete just the leaf node in the ZK but we ignore about the branch nodes. This is ok for FlatLedgerManager, but for HierarchicalLedgerManagers, especially for LongHierarchicalLedgerManager, the number of internal nodes gets blown up over time and we would get into ZK capacity limitations. When ZK reaches the capacity limits, it will manifest in very severe performance and stability issues of cluster. So for HierarchicalLedgerManagers, when we delete a ledger we should optimistically recursively delete the parent znodes as well if they don’t have anymore child znodes.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)