OHDSI Home | Forums | Wiki | Github

Save Our Sisyphus Challenge Week 6: Network Execution

General
#1

This week we learned all about network execution. Jenna Reps, Jack Brewster, Anthony Sena and Chungsoo Kim. Both tutorial sessions walked us through the steps to setup and run Strategus for network execution. Strategus allows us to run the necessary studies (characterization, prediction and estimation) and generates csv results. It’s very quick and easy use and takes only 15 minutes to run!

We are encouraging all data partners interested in any of the four SOS challenge studies to run the JSON specifications created by Patrick Ryan in Strategus. The Strategus tutorial on how to install and execute can be found here. If you have any questions or run into any issues, please join our Network execution office hour this Friday May 5th at 9:00AM EST.

Network Execution Office Hours: https://teams.microsoft.com/l/meetup-join/19%3Ameeting_NTQ1YTE0NGUtOTBmMC00YjdmLWFmNWMtZDkzMTMyMDYzZDE2%40thread.v2/0?context={"Tid"%3A"a30f0094-9120-4aab-ba4c-e5509023b2d5"%2C"Oid"%3A"970c82e9-6375-4f28-b728-2a49957a017f"}

#2

@anthonysena , thanks for a productive office hours. Is this an appropriate place for continued discussion?

If so, are any of these options available?:

  1. Have Strategus support forcing the use of single DatabaseConnector version
  2. Pass alternate versions of connectionDetails based upon DatabaseConnector version
  3. Hand-edit sub-module(s) to force them to use the right combination of the above?

I ask because for Databricks, the JDBC driver has changed across versions, and so has the connectionString. Older versions (JDBC <= 2.6.22) require jdbc::spark//, and newer versions (JDBC >= 2.6.25) require jdbc::databricks//. The net effect is that I can connect to the database for the early steps in Strategus, but as soon as I get to the CohortGenerator module, I get a JDBC connection error. For the first Strategus steps, DatabaseConnector is allowing the jdbc::databricks// connections, but not jdbc:://spark ones. For CohortGenerator, it allows jdbc::spark//, but not jdbc:databricks//

Do other databases have a similar challenge with DatabaseConnector versions, or is this unique to Databricks?

#3

As status update, changing the renv.lock files for each of the StrategusInstantiatedModules to use DatabaseConnector version 6.2.0 got us past nearly all of the JDBC errors.

t