bump spark-sql version to 3.2.0

[lmsurpre]

its only a provided dependency, but there's a few vulnerabilities in the transitive dependencies for 3.1.0.

dependabot opened this for it in the past #2985 ...we closed but I forget why

Signed-off-by: Lee Surprenant lmsurpre@us.ibm.com

[prb112]

yeah - it may not work properly with our environment. If you've tested this. I'll approve, I'd rather not accept it until it's tested

[lmsurpre]

I ran the following local smoketest:

1. marked the spark and stocator dependencies as runtime (instead of provided)
2. deployed to a local test server
3. configured the server with an ibm-cos storageProvider that has "enableParquet": true1
4. loaded a db with 1690 practitioner resources,
5. ran an export with _outputFormat=application/fhir+parquet and _type=Practitioner

I noted this, but I think we had the same warning with 3.1.0 and I don’t think there’s much we could do about it:

Illegal reflective access by org.apache.spark.unsafe.Platform (file:/Users/lmsurpre/.m2/repository/org/apache/spark/spark-unsafe_2.12/3.2.0/spark-unsafe_2.12-3.2.0.jar) to constructor java.nio.DirectByteBuffer(long,int)

I also noted some odd logging behavior, but I'm not sure if thats a regression either...still investigating.

[prb112]

LGTM - discussed with Lee offline

[lmsurpre]

I confirmed that the odd logging behavior is not a regression (we get the same odd behavior with the current versions).
I opened #3070 for that one and I think we're good to merge this in the meantime.