Export data to Parquet (Hadoop - Cloudera Stack) | Scheduled job

[astone42]

We have a Hadoop cluster that's based on the Cloudera Stack (CDH 5.8.3) and we are using parquet file format to store the data.
We want to export processed data from Splunk directly to the parquet tables in the Hadoop Cluster.

Example, let's assume a table named user_sessions exists in the Hadoop cluster stored in parquet.
1. User sessions log files are pushed to splunk
2. Scheduled Splunk query process the log files and outputs them in a table format
3. The data from step 2 is appended to the user_sessions table in the Hadoop cluster.

A possible solution for step 3 is to create a splunk custom command that connects to Impala through pyodbc and writes the data using INSERT INTO. The bottleneck for that solution is the performance.

Any ideas/suggestions?

Thanks a lot in advance.

[rdagan_splunk]

Martin, you can use Impala and DB Connect
Impala (notice the details on DBX3):
https://answers.splunk.com/answers/489448/can-i-connect-to-impala-sql-engine-on-hadoop-from.html
DB Connect dbxoutput command:
http://docs.splunk.com/Documentation/DBX/latest/DeployDBX/Createandmanagedatabaseoutputs