nsivabalan opened a new pull request, #12938:
URL: https://github.com/apache/hudi/pull/12938

   ### Change Logs
   
   Bloom index data skew bucket index partitioner.
   
   Fixing bloom index to tackle data skews in write path.
   
   For bucketized bloom, fixing the way we deduce the buckets per file group.
   Fixing the parallelism value for sort partitioner w/ bloom index.
   [Within each filegroup, we were determining the 
parallelism](https://github.com/apache/hudi/blob/aeebfcfcec271e8ff8f37e7f7ef2418d386f4c76/hudi-client/hudi-spark-client/src/main/java/org/apache/hudi/index/bloom/BucketizedBloomCheckPartitioner.java#L86)
 as Math.min(targetParallelism, totalBuckets). In this case, if incoming 
dataframe has 1 just spark partition, even though we could have computed the 
required buckets to be 10 (for eg), we might confine to just 1 bucket due to 
Math.min(). hence, resulting in 1 spark partition dealing with too many keys to 
be compared against. So, fixing this with this patch.
   
   Fix: If the bloom index parallelism is dynamically derived, we can go with 
Math.max instead of Math.min. This will ensure, we honor the total computed 
buckets instead of relying on dynamically derived parallelism from higher layer.
   But if the bloom index parallelism is explicitly overridden by the user, we 
can leave it as Math.min.
   
   ### Impact
   
   No data skews with bucketized partitioner with bloom index. 
   
   ### Risk level (write none, low medium or high below)
   
   low
   
   ### Documentation Update
   
   _Describe any necessary documentation update if there is any new feature, 
config, or user-facing change. If not, put "none"._
   
   - _The config description must be updated if new configs are added or the 
default value of the configs are changed_
   - _Any new feature or user-facing change requires updating the Hudi website. 
Please create a Jira ticket, attach the
     ticket number here and follow the 
[instruction](https://hudi.apache.org/contribute/developer-setup#website) to 
make
     changes to the website._
   
   ### Contributor's checklist
   
   - [ ] Read through [contributor's 
guide](https://hudi.apache.org/contribute/how-to-contribute)
   - [ ] Change Logs and Impact were stated clearly
   - [ ] Adequate tests were added if applicable
   - [ ] CI passed
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to