Skip to content

Deploying InfoConnect Hub with Docker, Kubernetes and Cloud Run

Overview

The InfoConnect Hub distribution archive includes ready-to-use deployment assets for container platforms. They are in the start-scripts/deploy folder and use the pre-built JARs from the info-connect-*/executable-jars/ folders, so no source code or Maven build is required. You need Docker, kubectl, a container registry and the cloud CLI of your platform (az for Azure, gcloud for Google Cloud).

This page is an overview. Each platform folder contains its own README or guide with the complete procedure, and each section below names the file to read. For a native installation without containers, see Installation, Services and Updates.

Target Folder in the distribution Start here
Docker, Hub only start-scripts/deploy/docker/hub/ README.md in the same folder
Docker, RPA Hub only start-scripts/deploy/docker/rpa/ README.md in the same folder
Docker, full stack (Hub, RPA, UI, MCP) start-scripts/deploy/docker/stack/ README.md in the same folder
Kubernetes (AKS, GKE, EKS, on-premises) start-scripts/deploy/k8s/ QUICKSTART.md and README.md in the same folder
Azure AKS with Azure Container Registry start-scripts/deploy/azure/ GUIDE.md in the same folder
Google GKE with Artifact Registry start-scripts/deploy/gcp/ GUIDE.md in the same folder
Google Cloud Run start-scripts/deploy/cloudrun/ QUICKSTART.md and GUIDE.md in the same folder

The files start-scripts/deploy/README.md and start-scripts/deploy/QUICKSTART.md index all platforms.

Note

Some READMEs inside the distribution show paths relative to the deploy folder (for example docker/hub). In the distribution archive, these folders are under start-scripts/deploy/.


Common prerequisites and files

License and truststore

The license file and the TLS truststore for IBM i are not included in the archive. Copy them into the security/ folder of the deployment you use:

Deployment Folder
Docker, Hub only start-scripts/deploy/docker/hub/security/ (truststore for IBM i connections)
Docker, RPA only start-scripts/deploy/docker/rpa/security/ (truststore for IBM i over TLS)
Docker, full stack start-scripts/deploy/docker/stack/security/
Kubernetes start-scripts/deploy/k8s/security/
Cloud Run A Google Cloud Storage bucket (SECURITY_BUCKET), see Google Cloud Run

The Kubernetes and Cloud Run guides expect two files with these names: as400-license.lic and info400new.truststore. In Docker and Kubernetes, the files are mounted read-only at /app/security inside the container. Do not commit the files or their passwords to source control.

A connection definition refers to the container path of the license or truststore, for example "filePath": "/app/security" in transportConfig.

Metadata database

The Hub keeps its metadata (connections, API definitions and DQ pollers) in a metadata database. The shipped READMEs describe these options:

Option Where it is described
H2, embedded (default) Docker, Hub only. Data persists in the Docker volume hub-config; failed records go to hub-data
JDBC (MySQL, PostgreSQL and other engines) Docker, Hub only; Docker, RPA only; Docker stack (MySQL); Kubernetes and Cloud Run (MySQL)
IBM i (libraries INFOCONNECT and INFOCONNECT_RPA) Docker, all variants; Kubernetes; Cloud Run

For IBM i and JDBC databases, create the schema and apply the DDL before the first start. The RPA Hub Docker setup uses spring.jpa.hibernate.ddl-auto=none, so its tables must exist before the first start.

Database DDL location in the distribution
IBM i, Hub (library INFOCONNECT) start-scripts/deploy/docker/hub/database_script/01-infoconnect-schema.sql
IBM i, RPA Hub (library INFOCONNECT_RPA) start-scripts/deploy/docker/rpa/database_script/AS400_DB_script.sql
MySQL start-scripts/deploy/docker/stack/mysql/init/

Configuration files

Deployment Copy To What to set
Docker, RPA only .env.example .env RPA_DATASOURCE_*, passwords, ENCRYPTION_WORD, truststore settings
Docker, full stack .env.example .env ENCRYPTION_WORD, MYSQL_* or AS400_*
Kubernetes env.example env.local IMAGE_REGISTRY, AS400_HOST, image tags, replicas
Kubernetes secrets.env.example secrets.env ENCRYPTION_WORD, AS400_PASSWORD, HUB_TRUSTSTORE_PASSWORD
Cloud Run env.example env.local and secrets.env GCP_PROJECT, GCP_REGION, AS400_HOST, image tags; ENCRYPTION_WORD, AS400_PASSWORD, HUB_TRUSTSTORE_PASSWORD

ENCRYPTION_WORD is the encryption key that the Hub uses to decrypt the connection passwords and truststore secrets sent in API requests. Use the same value for all services in a deployment.

Warning

The example files contain placeholder credentials. Before the first production deployment, replace the default Hub, UI and RPA user names and passwords and all values in the secrets file. Do not commit or share files that contain passwords (.env, secrets.env, a generated docker-compose.yml).


Docker, Hub only

Folder: start-scripts/deploy/docker/hub/. This deployment runs the Hub alone from the JAR in info-connect-hub/executable-jars/. The complete procedure is in start-scripts/deploy/docker/hub/README.md in the distribution.

  1. Prepare the metadata database. H2 needs no preparation. IBM i and JDBC need the schema and DDL (see Metadata database).
  2. For IBM i, copy the TLS truststore into security/.
  3. Run the configuration wizard:

    cd start-scripts/deploy/docker/hub
    
    ./configure-hub-docker.sh
    

    The wizard asks for the metadata database (Enter = H2, 1 = JDBC, 2 = IBM i), the Hub authentication, the encryption key, an optional RPA host and, for IBM i, the truststore settings. It writes docker-compose.yml.

  4. Start the Hub:

    ./run-hub.sh
    
  5. Verify that the Hub is running:

    curl -fsS http://localhost:8080/actuator/health
    

On Windows, run configure-hub-docker.bat (requires bash in the PATH). As an alternative to the wizard, copy docker-compose.yml.template, replace every __PLACEHOLDER__ and save the file as docker-compose.yml.

run-hub.sh (and run-hub.bat on Windows) also accepts stop, restart, logs and status. After you change settings, run the wizard again and then ./run-hub.sh restart.


Docker, RPA Hub only

Folder: start-scripts/deploy/docker/rpa/. This deployment runs the RPA Hub alone from the JAR in info-connect-rpa-hub/executable-jars/. The RPA metadata database can be MySQL, IBM i or any other JDBC database, configured with the RPA_DATASOURCE_* keys in .env. The complete procedure is in start-scripts/deploy/docker/rpa/README.md in the distribution.

Change to the folder and create the .env file:

cd start-scripts/deploy/docker/rpa
cp .env.example .env

Edit .env, then start the RPA Hub. With MySQL or another JDBC database:

docker compose up -d --build

With IBM i over TLS (set RPA_TRUSTSTORE_FILE and RPA_TRUSTSTORE_PASSWORD in .env and place the truststore in security/):

docker compose -f docker-compose.yml -f docker-compose.ibmi.yml up -d --build

To verify, call an RPA Hub API with the RPA administrator credentials from .env:

curl -u <rpa-admin-user>:<rpa-admin-password> http://localhost:8082/admin/connections

After you replace the JARs, rebuild the image with docker compose up -d --build.


Docker, full stack

Folder: start-scripts/deploy/docker/stack/. This deployment builds and runs Hub, RPA Hub, UI and MCP together. The build context is the distribution root. The complete procedure is in start-scripts/deploy/docker/stack/README.md in the distribution.

Choose the Docker Compose file that matches your metadata database:

Compose file Use when
docker-compose.yml The metadata is on MySQL (an external server or the bundled container)
docker-compose.as400.yml The metadata is on IBM i (libraries INFOCONNECT and INFOCONNECT_RPA)

Change to the folder and create the .env file:

cd start-scripts/deploy/docker/stack
cp .env.example .env

Edit .env (ENCRYPTION_WORD and either MYSQL_* or AS400_*), then start the stack.

With an external MySQL server (MYSQL_HOST must be reachable from the containers):

docker compose up -d --build

With the bundled MySQL 8 container (set MYSQL_HOST=mysql; the DDL from mysql/init/ is applied when MySQL first starts):

docker compose --profile local-mysql up -d --build

With IBM i (apply the DDL first and place the license and truststore in security/):

docker compose -f docker-compose.as400.yml up -d --build

The multi-hub profile of the MySQL stack adds a second Hub instance on port 8089:

docker compose --profile multi-hub up -d --build

Kubernetes

Folder: start-scripts/deploy/k8s/. The deploy-k8s.sh script (deploy-k8s.bat on Windows) deploys Hub, RPA Hub, UI and MCP to any Kubernetes cluster (AKS, GKE, EKS or on-premises). The complete procedure is in start-scripts/deploy/k8s/QUICKSTART.md and README.md in the distribution.

Prerequisites

  • A Kubernetes cluster and a working kubectl. Creating the cluster and downloading the kubeconfig requires az or gcloud.
  • Docker, to build and push the images.
  • Container images are pushed to a registry that the cluster nodes can pull from.
  • Network access from the cluster to the IBM i server (AS400_HOST), for example, through a VPN or a private link.
  • The IBM i DDL, which you apply once if IBM i is the metadata database.
  • The license and truststore in start-scripts/deploy/k8s/security/. The deployment continues without them, but IBM i TLS connections fail until the Kubernetes secret that contains these files exists.

Configure

Change to the folder and create the configuration files:

cd start-scripts/deploy/k8s
cp env.example env.local
cp secrets.env.example secrets.env

Edit env.local (IMAGE_REGISTRY, AS400_HOST and the image tags, which must exist in IMAGE_REGISTRY) and secrets.env (ENCRYPTION_WORD, AS400_PASSWORD, HUB_TRUSTSTORE_PASSWORD).

Build, deploy and verify

Build and push the images, deploy them and check the status:

./deploy-k8s.sh build
./deploy-k8s.sh deploy
./deploy-k8s.sh status
Command Action
./deploy-k8s.sh deploy Renders the manifests, applies the secrets and manifests and waits for the rollout
./deploy-k8s.sh build [service] Builds the images from the executable-jars folders and pushes them to the registry
./deploy-k8s.sh status Shows the pods and the LoadBalancer addresses
./deploy-k8s.sh scale hub N [--save-env] Sets the number of Hub replicas. --save-env stores the value in env.local
./deploy-k8s.sh scale --show Shows the current replica counts
./deploy-k8s.sh secrets Updates the secrets only
./deploy-k8s.sh destroy Deletes the Kubernetes namespace (default infoconnect) and everything in it

The deployment provides these endpoints:

Service Access Check
Hub LoadBalancer, port 8080 curl http://<hub-lb-ip>:8080/actuator/health
UI LoadBalancer, port 8085 Open http://<ui-lb-ip>:8085
MCP LoadBalancer, port 8083 curl http://<mcp-lb-ip>:8083/actuator/health
RPA Hub ClusterIP, port 8082 Use kubectl port-forward svc/rpa 8082:8082 -n infoconnect and call /admin/connections on localhost:8082

Updating

Change Steps
New distribution archive Replace the executable-jars folders, rebuild the images, run ./deploy-k8s.sh deploy
New image tag only Edit the tags in env.local and run ./deploy-k8s.sh deploy
Passwords Edit secrets.env, run ./deploy-k8s.sh secrets, then restart the deployments with kubectl rollout restart

update-from-distro does not apply to Docker or Kubernetes deployments.

Pod placement

The Kubernetes manifests for Hub, RPA Hub and UI define required pod anti-affinity by node (kubernetes.io/hostname). The pods of these three services run on different nodes, so none of them is scheduled on a node that already runs one of the other two. The MCP deployment has no such rule.


Azure AKS and Google GKE

The Kubernetes scripts and manifests are the same on every provider. The provider guides add the cluster and registry setup.

Azure AKS with Azure Container Registry

Guide: start-scripts/deploy/azure/GUIDE.md in the distribution. Related files in the same folder: BUILD.md (build and push the images) and LOGGING.md (Container Insights and log queries).

The guide covers:

  • Creating the resource group, the Azure Container Registry and the AKS cluster with az, and allowing the cluster to pull from the registry with az aks update --attach-acr.
  • Pushing the images and deploying with deploy-k8s.sh.
  • Verifying the metadata synchronization between Hub replicas.
  • Scaling the Hub and the day-to-day operations: status, logs, updating pods and rollback.

Prerequisites on your machine: the Azure CLI (az), kubectl and docker.

Google GKE with Artifact Registry

Guide: start-scripts/deploy/gcp/GUIDE.md in the distribution.

The guide covers:

  • Creating the GKE cluster (Autopilot or Standard) and the Artifact Registry repository with gcloud, and granting the cluster nodes read access to the repository when image pulls fail with ImagePullBackOff.
  • Configuring env.local with the Artifact Registry path as IMAGE_REGISTRY and deploying with deploy-k8s.sh.
  • Verifying the metadata synchronization and scaling the Hub.

Prerequisites on your machine: the Google Cloud CLI (gcloud), kubectl and docker.

Multiple Hub replicas

Both guides describe the following behavior when you run more than one Hub replica:

  • Use a shared external metadata database (IBM i or MySQL), not a separate embedded H2 database per pod. The manifests enable metadata synchronization (DB_SYNC_ENABLED=true), so each Hub pod reloads changes to Connections, APIs and DQ pollers from the shared database about every 60 seconds. The Docker and Kubernetes assets set this explicitly. In the distribution archive, db.sync.enabled is false by default.
  • More replicas increase the capacity and availability of the Functional APIs. The UI, RPA Hub and MCP default to one replica.
  • The Hub has no leader election for data queue (DQ) pollers. Every Hub pod starts all pollers that have the status ACTIVE, so several replicas become independent consumers of the same queue, and messages can be split unpredictably or duplicated. The guides recommend one Hub replica for steady-state DQ polling.

Warning

Scaling down does not stop pollers gracefully. Excess pods are terminated and the poller status in the database stays ACTIVE. If DQ pollers are running, stop each poller on every Hub pod before you reduce the replica count. The procedure is in section 4.4 of the AKS and GKE guides.


Google Cloud Run

Folder: start-scripts/deploy/cloudrun/. This deployment runs Hub, RPA Hub, MCP and UI as managed Cloud Run services with public HTTPS URLs. No Kubernetes cluster or kubectl is needed. The complete procedure is in start-scripts/deploy/cloudrun/QUICKSTART.md and GUIDE.md in the distribution. BUILD.md describes how to build the images and roll out new tags.

Prerequisites

  • gcloud, authenticated to your Google Cloud project (gcloud auth login).
  • docker, python and envsubst (part of gettext).
  • The images in Artifact Registry.
  • The IBM i DDL, which you apply once if IBM i is the metadata database.
  • The license and truststore files.

Configure and deploy

Change to the folder and create the configuration files:

cd start-scripts/deploy/cloudrun
cp env.example env.local
cp env.example secrets.env

Edit env.local (GCP_PROJECT, GCP_REGION, AS400_HOST, image tags) and secrets.env (ENCRYPTION_WORD, AS400_PASSWORD, HUB_TRUSTSTORE_PASSWORD). Then run the one-time setup. It creates the service account, the Secret Manager secrets and the Cloud Storage buckets:

./setup-gcp.sh

Upload the license and truststore to the security bucket (mounted at /app/security in Hub and RPA Hub). First load the variables:

source env.local

Then upload the files:

gsutil cp ../docker/stack/security/as400-license.lic "gs://${SECURITY_BUCKET}/"
gsutil cp ../docker/stack/security/info400new.truststore "gs://${SECURITY_BUCKET}/"

If the UI calls Hub and RPA Hub over HTTPS (*.run.app), run ./prepare-truststore.sh before you upload the truststore. The script adds the Google certificate authorities to the truststore.

Deploy all services (in the order RPA Hub, Hub, MCP, UI):

./deploy-cloudrun.sh deploy

To verify, check the health endpoint:

curl -fsS "${HUB_URL}/actuator/health"

HUB_URL is the URL of the infoconnect-hub service. To get it, run gcloud run services describe infoconnect-hub --region="${GCP_REGION}" --format='value(status.url)'.

Scaling

Cloud Run scales between a minimum and a maximum number of instances. By default, the Hub autoscales between 1 and 3 instances based on CPU load, while RPA Hub, MCP and UI stay at one instance. To set the Hub instance count:

./deploy-cloudrun.sh scale hub 3 --save-env
./deploy-cloudrun.sh scale --show

The scale command sets the minimum and the maximum to the same number. The DQ poller behavior described for Kubernetes replicas applies to Cloud Run instances as well. If you use DQ pollers, pin the Hub to a single instance (minimum and maximum of 1).

Other options

The guide also describes these options:

  • By default, the services are publicly reachable over HTTPS and protected by the InfoConnect credentials. You can restrict access with Google Cloud IAM instead (grant-invoker.sh).
  • If the IBM i server does not accept connections from Google Cloud public addresses, use a Serverless VPC Access connector (VPC_CONNECTOR in env.local).
  • To use MySQL instead of IBM i for metadata, replace the SPRING_DATASOURCE_* settings in services/hub-service.yaml.template and services/rpa-service.yaml.template with the MySQL JDBC settings.

Runtime changes in the deployment assets

Area Change
Docker images The full-stack Dockerfiles in start-scripts/deploy/docker/stack/ use an Alpine-based Java 17 runtime image (eclipse-temurin:17-jre-alpine)
RPA Hub The RPA Hub uses Jython 2.7.4 in the shipped Docker, Kubernetes and Cloud Run assets
RPA Hub TLS on Kubernetes The RPA Hub deployment pins TLS to version 1.2 and mounts a TLS security properties file (rpa-tls.security) from its ConfigMap
Pod placement on Kubernetes The manifests spread Hub, RPA Hub and UI pods across different nodes