FHIR Chat · Terminology Validators for Inferno Program Edition

Stream: inferno

Topic: Terminology Validators for Inferno Program Edition

Cooper Thompson (Jun 19 2020 at 19:21):

I'm working on generating the terminology validators for a locally deployed program edition. I've run the steps, but I'm trying to understand how the terminology container hooks up to the main docker containers. Does the terminology docker container generate an SQLite database that I copy over into the main docker containers for the program edition? Or does it host a terminology server that the main docker containers connect to in order to do validation at run-time?

Robert Scanlon (Jun 22 2020 at 14:13):

The output of the terminology process is only a number of files that get put in {INFERNO_ROOT}/resources/terminology/validators/. Technically we are generating a whole bunch of bloom filters (one per ValueSet and CodeSystem) that lets us do a set inclusion check. If that directory is empty then the process didn't work (let us know what you are seeing if that is the case and we'll help you figure out what went wrong). If that directory contains files, then it should automatically be mounted as a volume in inferno's docker-compose file. See https://github.com/onc-healthit/inferno-program/blob/master/docker-compose.yml#L9-L11

Robert Scanlon (Jun 22 2020 at 14:15):

FYI @Reece Adamson @Michael O'Keefe .

Robert Scanlon (Jun 22 2020 at 14:27):

One thing that is tripping a few people up is in v1.0.0 we did a poor job of naming the docker container that generates the terminology within the terminology process, and it will run with the same container name as the main inferno application. That was not intentional and we have a fix for that in the next release. If something gets stuck it may be because of that conflict (we can show you how to fix manually if needed).

Michael O'Keefe (Jun 22 2020 at 14:50):

Hi @Cooper Thompson , I'll follow this thread too. Let me know if you run into any issues with the terminology building process, or have any questions!

Cooper Thompson (Jul 08 2020 at 20:17):

I did end up getting the terminology files to generate. Just in case folks are interested, since the terminology generation container needs such a beefy server, I ended up running the terminology file generation on a beefy AWS VM, then copying the output files to the directory on a different server where I'm hosting the program edition website container. I did have a few issues on v0.9.6 that I never really investigated. My success was using v1.0.2.

Robert Scanlon (Jul 08 2020 at 20:52):

Thanks for sharing your experience Cooper. We are considering alternatives to this approach because having everyone regenerate valuesets is an awful waste of compute time.

Robert Scanlon (Jul 08 2020 at 20:58):

Would you mind sharing the manifest.yml file that was generated so we can double check that the right number of codes were generated? Like this: https://gist.github.com/arscan/3b43e24746762afada9b77dd5ec1323f

Cooper Thompson (Jul 09 2020 at 13:33):

manifest.yml

Cooper Thompson (Jul 09 2020 at 13:37):

Honestly, the process itself wasn't that complicated. The main issue I kept running into was that I was always running out of disk on the VM, so I kept having to expand and re-run, and I was always underestimating what I needed. I'd go from 8 to 16 to 32 to 64 GB of storage. Once I got a VM with the specs necessary, it was pretty smooth. For reference, my VM was a t2.xlarge Ubuntu 18.04 with 64GB of storage and 16 GB of RAM.

vency menezes (Aug 11 2020 at 06:18):

Cooper Thompson said: