[jira] [Closed] (HUDI-9039) Unneeded lock contention

["Danny Chen (Jira)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/HUDI-9039?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Danny Chen closed HUDI-9039.
----------------------------
     Reviewers: Danny Chen
    Resolution: Fixed

Fixed via master branch: 7103e50db7f6117d609c8bc6bac2fc777329a80a

> Unneeded lock contention
> ------------------------
>
>                 Key: HUDI-9039
>                 URL: https://issues.apache.org/jira/browse/HUDI-9039
>             Project: Apache Hudi
>          Issue Type: Bug
>            Reporter: Shawn Chang
>            Priority: Major
>              Labels: pull-request-available
>
> An EMR Hudi user reported that when upgrading from 0.8 to 0.15, with their existing multi-writer setup, they saw many lock contentions like below that they didn't have in 0.8.0:
>  
> {code:java}
> org.apache.hudi.exception.HoodieLockException: Unable to acquire lock, lock object null
>     at org.apache.hudi.client.transaction.lock.LockManager.lock(LockManager.java:91)
>     at org.apache.hudi.client.transaction.TransactionManager.beginTransaction(TransactionManager.java:59)
>     at org.apache.hudi.client.BaseHoodieWriteClient.doInitTable(BaseHoodieWriteClient.java:1248)
>     at org.apache.hudi.client.BaseHoodieWriteClient.initTable(BaseHoodieWriteClient.java:1291)
>     at org.apache.hudi.client.SparkRDDWriteClient.upsert(SparkRDDWriteClient.java:154) {code}
> This is likely because of BaseHoodieWriteClient in Hudi tries upgrade and init MDT for every write, even when upgrade is not needed or MDT is not enabled. See LOC: [https://github.com/apache/hudi/blob/f9b26dad4fcbf9d026aa31ae10ffa51a0db37935/hudi-client/hudi-client-common/src/main/java/org/apache/hudi/client/BaseHoodieWriteClient.java#L1302]
>  
> By looking at the code, this could impact the latest Hudi 1.0 as well
>  



--
This message was sent by Atlassian Jira
(v8.20.10#820010)