[ https://issues.apache.org/jira/browse/YARN-11183?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17633445#comment-17633445 ]
ASF GitHub Bot commented on YARN-11183: --------------------------------------- zhengchenyu commented on PR #4450: URL: https://github.com/apache/hadoop/pull/4450#issuecomment-1312944557 > @zhengchenyu Thank you very much for your contribution! From my personal point of view, I think adding AsyncDispatcher and defining a state machine to implement functions is a bit over-engineered. > > @goiri Thank you very much for inviting me to help review the code! > > I read this pr code carefully and I found that the function described by this pr has been implemented, The PRs involved are as follows: 1.[YARN-11290](https://issues.apache.org/jira/browse/YARN-11290)(#4846) 2.[YARN-11323](https://issues.apache.org/jira/browse/YARN-11323)(#4954) > > In [YARN-11290](https://issues.apache.org/jira/browse/YARN-11290), we improved the `getApplicationsHomeSubCluster` method, mainly in 2 improvement points: > > 1. Allows to query the App based on the specified HomeSubCluster. > 2. Currently this interface returns the number of apps in all sub-clusters, Added new configuration > used to limit Query the number of apps, return 1000 apps by default. > > In [YARN-11323](https://issues.apache.org/jira/browse/YARN-11323), we optimized the method of cleaning expired application data. More detailed information can be found in (#4954) > > From my personal point of view, I think [YARN-11290](https://issues.apache.org/jira/browse/YARN-11290) and [YARN-11323](https://issues.apache.org/jira/browse/YARN-11323) are enough, welcome to continue the discussion. @slfan1989 I read the PR https://github.com/apache/hadoop/pull/4954, it is lightweight way. But Why I not choose this way? removeApplicationIdFromStateStore(removeId) will connect to zookeeper. If zookeeper degrade, complete application will stuck. I think any operation which connect to 3rd store must do in async mode, It is why the statemachine of ZKRMStateStore is introduced. @goiri Welcome to give us some suggestion. Note: By the way, this PR is running on our cluster more than half a year. The number of znode in zookeeper keep steady. > Federation: Remove outdated ApplicationHomeSubCluster in federation state > store. > -------------------------------------------------------------------------------- > > Key: YARN-11183 > URL: https://issues.apache.org/jira/browse/YARN-11183 > Project: Hadoop YARN > Issue Type: Improvement > Components: federation, yarn > Reporter: zhengchenyu > Assignee: zhengchenyu > Priority: Major > Labels: pull-request-available > Time Spent: 1h > Remaining Estimate: 0h > > Nowadays, ApplicationHomeSubCluster in federation state store can't be > removed automatically. -- This message was sent by Atlassian Jira (v8.20.10#820010) --------------------------------------------------------------------- To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org