Skip to content

Latest commit

 

History

History
599 lines (436 loc) · 26.2 KB

README.md

File metadata and controls

599 lines (436 loc) · 26.2 KB
Raw

HAPI-FHIR Starter Project

This project is a complete starter project you can use to deploy a FHIR server using HAPI FHIR JPA.

Note that this project is specifically intended for end users of the HAPI FHIR JPA server module (in other words, it helps you implement HAPI FHIR, it is not the source of the library itself). If you are looking for the main HAPI FHIR project, see here: https://github.com/hapifhir/hapi-fhir

While this project shows how you can use many parts of the HAPI FHIR framework there are a set of features which you should be aware of are missing or something you need to supply yourself or get professional support ahead of using it directly in production:

  1. The service comes with no security implementation. See how it can be done here
  2. The service comes with no enterprise logging. See how it can be done here
  3. The internal topic cache used by subscriptions in HAPI FHIR are not shared across multiple instances as the default supplied implementation is in-mem
  4. The internal message broker channel in HAPI FHIR is not shared across multiple instances as the default supplied implementation is in-mem. This impacts the use of modules listed here

Need Help? Please see: https://github.com/hapifhir/hapi-fhir/wiki/Getting-Help

Prerequisites

In order to use this sample, you should have:

  • This project checked out. You may wish to create a GitHub Fork of the project and check that out instead so that you can customize the project and save the results to GitHub.

and either

  • Oracle Java (JDK) installed: Minimum JDK17 or newer.
  • Apache Maven build tool (newest version)

or

  • Docker, as the entire project can be built using multistage docker (with both JDK and maven wrapped in docker) or used directly from Docker Hub

Running via Docker Hub

Each tagged/released version of hapi-fhir-jpaserver is built as a Docker image and published to Docker hub. To run the published Docker image from DockerHub:

docker pull hapiproject/hapi:latest
docker run -p 8080:8080 hapiproject/hapi:latest

This will run the docker image with the default configuration, mapping port 8080 from the container to port 8080 in the host. Once running, you can access http://localhost:8080/ in the browser to access the HAPI FHIR server's UI or use http://localhost:8080/fhir/ as the base URL for your REST requests.

If you change the mapped port, you need to change the configuration used by HAPI to have the correct hapi.fhir.tester property/value.

Configuration via environment variables

You can customize HAPI directly from the run command using environment variables. For example:

docker run -p 8080:8080 -e hapi.fhir.default_encoding=xml hapiproject/hapi:latest

HAPI looks in the environment variables for properties in the application.yaml file for defaults.

Configuration via overridden application.yaml file and using Docker

You can customize HAPI by telling HAPI to look for the configuration file in a different location, eg.:

docker run -p 8090:8080 -v $(pwd)/yourLocalFolder:/configs -e "--spring.config.location=file:///configs/another.application.yaml" hapiproject/hapi:latest

Here, the configuration file (another.application.yaml) is placed locally in the folder yourLocalFolder.

docker run -p 8090:8080 -e "--spring.config.location=classpath:/another.application.yaml" hapiproject/hapi:latest

Here, the configuration file (another.application.yaml) is part of the compiled set of resources.

Example using docker-compose.yml for docker-compose

version: '3.7'

services:
  fhir:
    container_name: fhir
    image: "hapiproject/hapi:latest"
    ports:
      - "8080:8080"
    configs:
      - source: hapi
        target: /app/config/application.yaml
    depends_on:
      - db


  db:
    image: postgres
    restart: always
    environment:
      POSTGRES_PASSWORD: admin
      POSTGRES_USER: admin
      POSTGRES_DB: hapi
    volumes:
      - ./hapi.postgress.data:/var/lib/postgresql/data

configs:
  hapi:
     file: ./hapi.application.yaml

Provide the following content in ./hapi.application.yaml:

spring:
  datasource:
    url: 'jdbc:postgresql://db:5432/hapi'
    username: admin
    password: admin
    driverClassName: org.postgresql.Driver
  jpa:
    properties:
      hibernate.dialect: ca.uhn.fhir.jpa.model.dialect.HapiFhirPostgresDialect
      hibernate.search.enabled: false

Example running custom interceptor using docker-compose

This example is an extension of the above one, now adding a custom interceptor.

version: '3.7'

services:
  fhir:
    container_name: fhir
    image: "hapiproject/hapi:latest"
    ports:
      - "8080:8080"
    configs:
      - source: hapi
        target: /app/config/application.yaml
      - source: hapi-extra-classes
        target: /app/extra-classes
    depends_on:
      - db

  db:
    image: postgres
    restart: always
    environment:
      POSTGRES_PASSWORD: admin
      POSTGRES_USER: admin
      POSTGRES_DB: hapi
    volumes:
      - ./hapi.postgress.data:/var/lib/postgresql/data

configs:
  hapi:
     file: ./hapi.application.yaml
  hapi-extra-classes:
     file: ./hapi-extra-classes

Provide the following content in ./hapi.application.yaml:

spring:
  datasource:
    url: 'jdbc:postgresql://db:5432/hapi'
    username: admin
    password: admin
    driverClassName: org.postgresql.Driver
  jpa:
    properties:
      hibernate.dialect: ca.uhn.fhir.jpa.model.dialect.HapiFhirPostgresDialect
      hibernate.search.enabled: false
hapi:
  fhir:
    custom-bean-packages: the.package.containing.your.interceptor
    custom-interceptor-classes: the.package.containing.your.interceptor.YourInterceptor

The basic interceptor structure would be like this:

package the.package.containing.your.interceptor;

import org.hl7.fhir.instance.model.api.IBaseResource;
import org.springframework.stereotype.Component;

import ca.uhn.fhir.interceptor.api.Hook;
import ca.uhn.fhir.interceptor.api.Interceptor;
import ca.uhn.fhir.interceptor.api.Pointcut;

@Component
@Interceptor
public class YourInterceptor
{
    @Hook(Pointcut.STORAGE_PRECOMMIT_RESOURCE_CREATED)
    public void resourceCreated(IBaseResource newResource)
    {
        System.out.println("YourInterceptor.resourceCreated");
    }
}

Running locally

The easiest way to run this server entirely depends on your environment requirements. The following ways are supported:

Using jetty

mvn -Pjetty spring-boot:run

The Server will then be accessible at http://localhost:8080/fhir and the CapabilityStatement will be found at http://localhost:8080/fhir/metadata.

Using Spring Boot

mvn spring-boot:run

The Server will then be accessible at http://localhost:8080/fhir and the CapabilityStatement will be found at http://localhost:8080/fhir/metadata.

If you want to run this server on a different port, you can change the port in the src/main/resources/application.yaml file as follows:

server:
#  servlet:
#    context-path: /example/path
  port: 8888

The Server will then be accessible at http://localhost:8888/fhir and the CapabilityStatement will be found at http://localhost:8888/fhir/metadata. Remember to adjust your overlay configuration in the application.yaml file to the following:

    tester:
      -
          id: home
          name: Local Tester
          server_address: 'http://localhost:8888/fhir'
          refuse_to_fetch_third_party_urls: false
          fhir_version: R4

Using Spring Boot with :run

mvn clean spring-boot:run -Pboot

Server will then be accessible at http://localhost:8080/ and eg. http://localhost:8080/fhir/metadata. Remember to adjust you overlay configuration in the application.yaml to the following:

    tester:
      -
          id: home
          name: Local Tester
          server_address: 'http://localhost:8080/fhir'
          refuse_to_fetch_third_party_urls: false
          fhir_version: R4

Using Spring Boot

mvn clean package spring-boot:repackage -DskipTests=true -Pboot && java -jar target/ROOT.war

Server will then be accessible at http://localhost:8080/ and eg. http://localhost:8080/fhir/metadata. Remember to adjust your overlay configuration in the application.yaml to the following:

    tester:
      -
          id: home
          name: Local Tester
          server_address: 'http://localhost:8080/fhir'
          refuse_to_fetch_third_party_urls: false
          fhir_version: R4

Using Spring Boot and Google distroless

mvn clean package com.google.cloud.tools:jib-maven-plugin:dockerBuild -Dimage=distroless-hapi && docker run -p 8080:8080 distroless-hapi

Server will then be accessible at http://localhost:8080/ and eg. http://localhost:8080/fhir/metadata. Remember to adjust your overlay configuration in the application.yaml to the following:

    tester:
      -
          id: home
          name: Local Tester
          server_address: 'http://localhost:8080/fhir'
          refuse_to_fetch_third_party_urls: false
          fhir_version: R4

Using the Dockerfile and multistage build

./build-docker-image.sh && docker run -p 8080:8080 hapi-fhir/hapi-fhir-jpaserver-starter:latest

Server will then be accessible at http://localhost:8080/ and eg. http://localhost:8080/fhir/metadata. Remember to adjust your overlay configuration in the application.yaml to the following:

    tester:
      -
          id: home
          name: Local Tester
          server_address: 'http://localhost:8080/fhir'
          refuse_to_fetch_third_party_urls: false
          fhir_version: R4

You can use a custom property file that utilizes environment variables for many configuration properties. For example, to use the application-custom.yaml file. Replace with actual values.

./build-docker-image.sh && docker run -p 8080:8080 \
  -e spring.config.location='<classpath:/application-custom.yaml>' \
  hapi-fhir/hapi-fhir-jpaserver-starter:latest

Configurations

Much of this HAPI starter project can be configured using the yaml file in src/main/resources/application.yaml. By default, this starter project is configured to use H2 as the database.

MySQL configuration

HAPI FHIR JPA Server does not support MySQL as it is deprecated.

See more at https://hapifhir.io/hapi-fhir/docs/server_jpa/database_support.html

PostgreSQL configuration

To configure the starter app to use PostgreSQL, instead of the default H2, update the application.yaml file to have the following:

spring:
  datasource:
    url: 'jdbc:postgresql://localhost:5432/hapi'
    username: admin
    password: admin
    driverClassName: org.postgresql.Driver
  jpa:
    properties:
      hibernate.dialect: ca.uhn.fhir.jpa.model.dialect.HapiFhirPostgresDialect
      hibernate.search.enabled: false

      # Then comment all hibernate.search.backend.*

Because the integration tests within the project rely on the default H2 database configuration, it is important to either explicitly skip the integration tests during the build process, i.e., mvn install -DskipTests, or delete the tests altogether. Failure to skip or delete the tests once you've configured PostgreSQL for the datasource.driver, datasource.url, and hibernate.dialect as outlined above will result in build errors and compilation failure.

Microsoft SQL Server configuration

To configure the starter app to use MS SQL Server, instead of the default H2, update the application.yaml file to have the following:

spring:
  datasource:
    url: 'jdbc:sqlserver://<server>:<port>;databaseName=<databasename>'
    username: admin
    password: admin
    driverClassName: com.microsoft.sqlserver.jdbc.SQLServerDriver

Also, make sure you are not setting the Hibernate dialect explicitly, in other words, remove any lines similar to:

hibernate.dialect: {some none Microsoft SQL dialect}

Because the integration tests within the project rely on the default H2 database configuration, it is important to either explicitly skip the integration tests during the build process, i.e., mvn install -DskipTests, or delete the tests altogether. Failure to skip or delete the tests once you've configured PostgreSQL for the datasource.driver, datasource.url, and hibernate.dialect as outlined above will result in build errors and compilation failure.

NOTE: MS SQL Server by default uses a case-insensitive codepage. This will cause errors with some operations - such as when expanding case-sensitive valuesets (UCUM) as there are unique indexes defined on the terminology tables for codes. It is recommended to deploy a case-sensitive database prior to running HAPI FHIR when using MS SQL Server to avoid these and potentially other issues.

Adding custom interceptors

Custom interceptors can be registered with the server by including the property hapi.fhir.custom-interceptor-classes. This will take a comma separated list of fully-qualified class names which will be registered with the server. Interceptors will be discovered in one of two ways:

  1. discovered from the Spring application context as existing Beans (can be used in conjunction with hapi.fhir.custom-bean-packages) or registered with Spring via other methods

or

  1. classes will be instantiated via reflection if no matching Bean is found

Adding custom operations(providers)

Custom operations(providers) can be registered with the server by including the property hapi.fhir.custom-provider-classes. This will take a comma separated list of fully-qualified class names which will be registered with the server. Providers will be discovered in one of two ways:

  1. discovered from the Spring application context as existing Beans (can be used in conjunction with hapi.fhir.custom-bean-packages) or registered with Spring via other methods

or

  1. classes will be instantiated via reflection if no matching Bean is found

Customizing The Web Testpage UI

The UI that comes with this server is an exact clone of the server available at http://hapi.fhir.org. You may skin this UI if you'd like. For example, you might change the introductory text or replace the logo with your own.

The UI is customized using Thymeleaf template files. You might want to learn more about Thymeleaf, but you don't necessarily need to: they are quite easy to figure out.

Several template files that can be customized are found in the following directory: https://github.com/hapifhir/hapi-fhir-jpaserver-starter/tree/master/src/main/webapp/WEB-INF/templates

Deploying to an Application Server

Using the Maven-Embedded Jetty method above is convenient, but it is not a good solution if you want to leave the server running in the background.

Most people who are using HAPI FHIR JPA as a server that is accessible to other people (whether internally on your network or publically hosted) will do so using an Application Server, such as Apache Tomcat or Jetty. Note that any Servlet 3.0+ compatible Web Container will work (e.g Wildfly, Websphere, etc.).

Tomcat is very popular, so it is a good choice simply because you will be able to find many tutorials online. Jetty is a great alternative due to its fast startup time and good overall performance.

To deploy to a container, you should first build the project:

mvn clean install

This will create a file called ROOT.war in your target directory. This should be installed in your Web Container according to the instructions for your particular container. For example, if you are using Tomcat, you will want to copy this file to the webapps/ directory.

Again, browse to the following link to use the server (note that the port 8080 may not be correct depending on how your server is configured).

http://localhost:8080/

You will then be able to access the JPA server e.g. using http://localhost:8080/fhir/metadata.

If you would like it to be hosted at eg. hapi-fhir-jpaserver, eg. http://localhost:8080/hapi-fhir-jpaserver/ or http://localhost:8080/hapi-fhir-jpaserver/fhir/metadata - then rename the WAR file to hapi-fhir-jpaserver.war and adjust the overlay configuration accordingly e.g.

    tester:
      -
          id: home
          name: Local Tester
          server_address: 'http://localhost:8080/hapi-fhir-jpaserver/fhir'
          refuse_to_fetch_third_party_urls: false
          fhir_version: R4

Deploy with docker compose

Docker compose is a simple option to build and deploy containers. To deploy with docker compose, you should build the project with mvn clean install and then bring up the containers with docker-compose up -d --build. The server can be reached at http://localhost:8080/.

In order to use another port, change the ports parameter inside docker-compose.yml to 8888:8080, where 8888 is a port of your choice.

The docker compose set also includes PostgreSQL database, if you choose to use PostgreSQL instead of H2, change the following properties in src/main/resources/application.yaml:

spring:
  datasource:
    url: 'jdbc:postgresql://hapi-fhir-postgres:5432/hapi'
    username: admin
    password: admin
    driverClassName: org.postgresql.Driver
jpa:
  properties:
    hibernate.dialect: ca.uhn.fhir.jpa.model.dialect.HapiFhirPostgresDialect
    hibernate.search.enabled: false

    # Then comment all hibernate.search.backend.*

Running hapi-fhir-jpaserver directly from IntelliJ as Spring Boot

Make sure you run with the maven profile called boot and NOT also jetty. Then you are ready to press debug the project directly without any extra Application Servers.

Running hapi-fhir-jpaserver-example in Tomcat from IntelliJ

Install Tomcat.

Make sure you have Tomcat set up in IntelliJ.

  • File->Settings->Build, Execution, Deployment->Application Servers
  • Click +
  • Select "Tomcat Server"
  • Enter the path to your tomcat deployment for both Tomcat Home (IntelliJ will fill in base directory for you)

Add a Run Configuration for running hapi-fhir-jpaserver-example under Tomcat

  • Run->Edit Configurations
  • Click the green +
  • Select Tomcat Server, Local
  • Change the name to whatever you wish
  • Uncheck the "After launch" checkbox
  • On the "Deployment" tab, click the green +
  • Select "Artifact"
  • Select "hapi-fhir-jpaserver-example:war"
  • In "Application context" type /hapi

Run the configuration.

  • You should now have an "Application Servers" in the list of windows at the bottom.
  • Click it.
  • Select your server, and click the green triangle (or the bug if you want to debug)
  • Wait for the console output to stop

Point your browser (or fiddler, or what have you) to http://localhost:8080/hapi/baseDstu3/Patient

Enabling Subscriptions

The server may be configured with subscription support by enabling properties in the application.yaml file:

  • hapi.fhir.subscription.resthook_enabled - Enables REST Hook subscriptions, where the server will make an outgoing connection to a remote REST server

  • hapi.fhir.subscription.email.* - Enables email subscriptions. Note that you must also provide the connection details for a usable SMTP server.

  • hapi.fhir.subscription.websocket_enabled - Enables websocket subscriptions. With this enabled, your server will accept incoming websocket connections on the following URL (this example uses the default context path and port, you may need to tweak depending on your deployment environment): ws://localhost:8080/websocket

Enabling Clinical Reasoning

Set hapi.fhir.cr.enabled=true in the application.yaml file to enable Clinical Quality Language on this server. An alternate settings file, cds.application.yaml, exists with the Clinical Reasoning module enabled and default settings that have been found to work with most CDS and dQM test cases.

Enabling CDS Hooks

Set hapi.fhir.cdshooks.enabled=true in the application.yaml file to enable CDS Hooks on this server. The Clinical Reasoning module must also be enabled because this implementation of CDS Hooks includes CDS on FHIR. An example CDS Service using CDS on FHIR is available in the CdsHooksServletIT test class.

Enabling MDM (EMPI)

Set hapi.fhir.mdm_enabled=true in the application.yaml file to enable MDM on this server. The MDM matching rules are configured in mdm-rules.json. The rules in this example file should be replaced with actual matching rules appropriate to your data. Note that MDM relies on subscriptions, so for MDM to work, subscriptions must be enabled.

Using Elasticsearch

By default, the server will use embedded lucene indexes for terminology and fulltext indexing purposes. You can switch this to using lucene by editing the properties in application.yaml

For example:

elasticsearch.enabled=true
elasticsearch.rest_url=localhost:9200
elasticsearch.username=SomeUsername
elasticsearch.password=SomePassword
elasticsearch.protocol=http
elasticsearch.required_index_status=YELLOW
elasticsearch.schema_management_strategy=CREATE

Enabling LastN

Set hapi.fhir.lastn_enabled=true in the application.yaml file to enable the $lastn operation on this server. Note that the $lastn operation relies on Elasticsearch, so for $lastn to work, indexing must be enabled using Elasticsearch.

Enabling Resource to be stored in Lucene Index

Set hapi.fhir.store_resource_in_lucene_index_enabled in the application.yaml file to enable storing of resource json along with Lucene/Elasticsearch index mappings.

Changing cached search results time

It is possible to change the cached search results time. The option reuse_cached_search_results_millis in the application.yaml is 6000 miliseconds by default. Set reuse_cached_search_results_millis: -1 in the application.yaml file to ignore the cache time every search.

Build the distroless variant of the image (for lower footprint and improved security)

The default Dockerfile contains a release-distroless stage to build a variant of the image using the gcr.io/distroless/java-debian10:11 base image:

docker build --target=release-distroless -t hapi-fhir:distroless .

Note that distroless images are also automatically built and pushed to the container registry, see the -distroless suffix in the image tags.

Adding custom operations

To add a custom operation, refer to the documentation in the core hapi-fhir libraries here.

Within hapi-fhir-jpaserver-starter, create a generic class (that does not extend or implement any classes or interfaces), add the @Operation as a method within the generic class, and then register the class as a provider using RestfulServer.registerProvider().

Runtime package install

It's possible to install a FHIR Implementation Guide package (package.tgz) either from a published package or from a local package with the $install operation, without having to restart the server. This is available for R4 and R5.

This feature must be enabled in the application.yaml (or docker command line):

hapi:
  fhir:
    ig_runtime_upload_enabled: true

The $install operation is triggered with a POST to [server]/ImplementationGuide/$install, with the payload below:

{
  "resourceType": "Parameters",
  "parameter": [
    {
      "name": "npmContent",
      "valueBase64Binary": "[BASE64_ENCODED_NPM_PACKAGE_DATA]"
    }
  ]
}

Enable OpenTelemetry auto-instrumentation

The container image includes the OpenTelemetry Java auto-instrumentation Java agent JAR which can be used to export telemetry data for the HAPI FHIR JPA Server. You can enable it by specifying the -javaagent flag, for example by overriding the JAVA_TOOL_OPTIONS environment variable:

docker run --rm -it -p 8080:8080 \
  -e JAVA_TOOL_OPTIONS="-javaagent:/app/opentelemetry-javaagent.jar" \
  -e OTEL_TRACES_EXPORTER="jaeger" \
  -e OTEL_SERVICE_NAME="hapi-fhir-server" \
  -e OTEL_EXPORTER_JAEGER_ENDPOINT="http://jaeger:14250" \
  docker.io/hapiproject/hapi:latest

You can configure the agent using environment variables or Java system properties, see https://opentelemetry.io/docs/instrumentation/java/automatic/agent-config/ for details.