Glue job runs indefinitely without finishing code execution - pyspark

Question

I have a glue job that reads in data from an RDS postgres instance (via data catalog) and writes to s3 in partitioned and parquet format. That works fine. I added to the end of that script code to run a crawler on the s3 path that was written to, so that the data catalog has the new partitions updated. But this code never runs. I added a logging statement immediately after the line that writes the dynamic frame, and this logging statement is never output. The job just runs indefinitely until time out or I stop the run.
import ...
...
sc = SparkContext()
gluecontext = GlueContext(sc)
log4jLogger=sc._jvm.org.apache.log4j
log=log4jLogger.LogManager.getLogger(__name__)
log.warn("test of logger")
DyF = gluecontext.create_dynamic_frame_from_catalog(database='db-we-want', table_name='table-of-value')
create_paritions(DyF)
log.warn("about to write dynamic frame with data of interest")
gluecontext.write_dyanmic_frame_from_options(frame=DyF, connection_type='s3', connection_options={'path': 's3://some-bucket/some-prefix'}, format='parquet')
log.warn('Attempting to start crawler')
glue_client = boto3.client('glue', region_name='us-east-1')
glue_client.start_crawler(Name='some-crawler')
I expect to have the crawler start and to see the log statements after the first glue context writes. I see objects in s3 for the glue context dynamic frame, but the log statement immediately after doesn't write. The crawler does not start. There are no errors in the logs and the job continues to run indefinitely.
EDIT:
I was able to solve this issue. The problem was that the RDS connection subnet was public. However glue jobs do not have a public IP. They need access to an NAT gateway. By switching the connection to use a private subnet with an NAT gateway the job was able to succeed. The error was a boto3 time out on trying to connect to glue resources.