This message was deleted.
# general
s
This message was deleted.
s
I believe it supports zstd compression as mentioned here https://druid.apache.org/docs/latest/ingestion/native-batch.html#supported-compression-formats. Not sure if it is supported in 0.19 though.
How do we set the property to accept zstd compressed parquet data in druid server? What is the property and where do we set it?
c
I don't see a property where you specify the incoming compression
g
IIRC for Parquet decompression is automatic as part of the reader
s
Thanks Charles and Gian.
I have some follow-up questions: 1. What property is needed to be set on druid ingestion spec to ingest zstd compressed parquet data on druid version 0.19? 2. Do we need to install any packages on druid server for druid to parse the zstd compressed parquet data?
g
unfortunately I don't know the answers since I haven't use parquet with zstd before 😅
Also, Druid version 0.19 is really old 🙂
One path here could be to add a zstd Parquet file to https://github.com/apache/druid/blob/master/extensions-core/parquet-extensions/example/compat And add a test case in
CompatParquetReaderTest
This'll validate it works in the latest version, and in case there is some reason it doesn't, it would provide a test case to help figure out a fix
s
No problem. Maybe someone else from the community might know the answer. Yeah we are moving to 0.25 in a new cluster, but for now we are maintaining 0.19.
if the above way you mentioned can be done, that would be nice. Do we need to open a pull request for the same?
g
If you do; it'd be greatly appreciated and would help ensure it always works going forward!
more test cases are always a good thing
If you do, post the PR link here and I'll have a look
s
ok, I mostly look at the infra side, I am not that comfortable with it. Let me check with other team member if he knows how to do it.
@dontu balu will you be able to try it out?