Install OpenSourceOM on Kubernetes

The chart in deploy/helm/opensourceom installs the API Deployment and, by default, a Postgres StatefulSet. Chart version and appVersion are 0.3.0. When bundled Postgres is enabled, an init container named wait-db runs pg_isready against {{release}}-opensourceom-postgres. A second init container runs om migrate. The main container runs om serve. Readiness and liveness probe GET /v1/health. With an external database, wait-db is omitted and migrate still runs, so that host must accept connections before the pod can become Ready.

Install

api.secret and postgres.password are required. The image repository defaults to ghcr.io/opensourceom/core. An empty image.tag falls back to the chart appVersion (0.3.0). Set the tag when you are not running that release.

docker build -t ghcr.io/opensourceom/core:0.3.0 .
helm install om deploy/helm/opensourceom \
  --set api.secret='change-me' \
  --set postgres.password='change-me' \
  --set image.tag=0.3.0

Load that image into the cluster if you are not pushing to GHCR. Published images live at ghcr.io/opensourceom/core. The Service is ClusterIP on port 8080. The release name om yields a Service named om-opensourceom:

kubectl port-forward svc/om-opensourceom 8080:8080

Values

Value	Default	Role
`replicaCount`	`1`	API replicas. They share one database.
`image.repository`	`ghcr.io/opensourceom/core`	API image.
`image.tag`	empty, then chart `appVersion` (`0.3.0`)	Image tag. Set it when the cluster should not run `0.3.0`.
`service.type` / `service.port`	`ClusterIP` / `8080`	How the API is reached inside the cluster.
`api.env`	`production`	Written to `OM_ENV`.
`api.secret`	empty, required	`OM_API_SECRET`.
`postgres.enabled`	`true`	When false, set `postgres.host` to your own server.
`postgres.persistence.size`	`10Gi`	PVC for the bundled database. `storageClass` is optional.
`resources`	100m / 128Mi request, 1 CPU / 512Mi limit	API container. Scan CronJobs use the same limits unless `scan.resources` is set.
`scan.schedule`	`0 /6 * *`	Schedule for every enabled collector. Overlapping runs of one collector are skipped.

Scheduled scans

Collectors stay off until you enable one and supply credentials. The chart then runs om scan as a CronJob against the same Postgres the API reads. For a release named om, that Service is om-opensourceom-postgres. om scan demo is not scheduled. A scan does not run CSPM rules.

helm upgrade om deploy/helm/opensourceom \
  --reuse-values \
  --set scan.aws.enabled=true \
  --set scan.aws.existingSecret=aws-creds

aws-creds must contain AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY. Azure needs scan.azure.subscriptionId and a Secret with AZURE_TENANT_ID, AZURE_CLIENT_ID, and AZURE_CLIENT_SECRET. GCP needs scan.gcp.projectId and a Secret key named credentials.json. Inline key fields on scan.aws, scan.azure, and scan.gcp work the same way and are copied into a chart Secret. serviceAccountAuth: true skips the static key and uses the chart ServiceAccount; set scan.serviceAccount.annotations for IRSA or workload identity. Enabling a cloud collector without one of those credentials does not create a CronJob.

scan.kubernetes.enabled=true schedules om scan k8s. With no kubeconfig Secret, the chart creates a read-only ServiceAccount for this cluster. scan.kubernetes.namespace empty is a ClusterRole; a namespace is a Role in that namespace. scan.kubernetes.cluster is the graph account id. Set scan.kubernetes.existingSecret to mount a kubeconfig instead (key kubeconfig, unless kubeconfigKey says otherwise). See the Kubernetes collector.

Set postgres.enabled=false and postgres.host to use a database you already run. postgres.host is required in that mode. The chart still creates the API Secret. It does not create a database user on the external server.

Keep the Service off the public internet, or put your own gateway in front of it. See Security for how OM_API_SECRET applies.