koertkuipers commented on issue #25739: [WIP][SPARK-28945][CORE][SQL] Support concurrent dynamic partition writes to different partitions in the same table URL: https://github.com/apache/spark/pull/25739#issuecomment-529969738 > I labeled this as [WIP] as I think we can also enable concurrent writes to the same table with dynamicPartitionOverwrite disabled. When we are writing to table dynamically in the strict mode, we can support concurrent writes to different static partitions. However that would require more changes and I'd like to know others opinions. > > cc @cloud-fan @koertkuipers what would concurrent writes to the same table with dynamicPartitionOverwrite disabled look like? i have a hard time coming up with a useful example of this. > > what would concurrent writes to the same table with dynamicPartitionOverwrite disabled look like? i have a hard time coming up with a useful example of this. > > Suppose we have a table with partition columns (day, hour, action). It would be useful to support > concurrent writes to `Partition(day=20190910, hour=01, action)`, `Partition(day=20190910, hour=02, action)` and etc. Or to Partitions: `Partition(day=20190910, hour, action)`, `Partition(day=20190909, hour, action)` and etc. > > The concurrent write would be succeeded as long as the static partitions share the same size. my understanding is that currently if you don't have dynamic partition overwrite enabled it will always delete all partitions before writing. i don't see concurrency being useful in this situation. the example you give sounds interesting to me but its inconsistent with how i currently know static partition overwrite to function. would this be a new feature?
---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
