Skip to content

Fortified IDDocs

Container

Run Fortified ID Signing Server in Docker with Compose, an external or bundled PostgreSQL, and protected configuration mounts.

This page describes how to run Signing Server as a container on Linux x86_64. The image contains the same archive as the installers.

  • Docker with Compose, and the Signing Server image. Reference the image by its immutable ID or registry digest in SIGNING_IMAGE.
  • A configuration directory containing server.json, the key stores, the identity provider configuration and the referenced files.
  • A directory with the database password files.

The container runs as the numeric user and group 10001:10001. The runtime does not change ownership, generate keys or repair a database. Configuration, key and password files must be regular files with mode 0600 in directories with mode 0700.

Create an environment file that contains paths only, never secrets:

SIGNING_IMAGE=sha256:REPLACE_WITH_IMAGE_ID
SIGNING_CONFIGURATION_DIR=/srv/signing/config
SIGNING_DATABASE_SECRETS_DIR=/srv/signing/database-passwords
SIGNING_PUBLISH_ADDRESS=127.0.0.1
SIGNING_PUBLISH_PORT=8443

The configuration and password mounts are read-only. Paths in server.json refer to the mount points /etc/signing for configuration and /run/signing-db for database passwords.

Create the named volumes work and logs, for a project named signing as signing_work and signing_logs. Set each volume root to owner 10001:10001 and mode 0700 before the first start. All named mounts use nocopy, so Docker never copies image permissions onto an existing volume, and a volume with wrong ownership makes startup fail.

Set the database host, port and name in server.json, and use ssl_mode verify-full with the CA file for a production database. Accounts and privileges are provisioned by the database administrator.

Terminal window
docker compose -p signing --env-file /srv/signing/deployment.env \
-f deploy/docker/compose.yaml up -d --wait

For an empty new installation you can add a PostgreSQL service with compose.postgresql.yaml. It runs as 10001:10001 with a read-only root filesystem and no published port.

Add these references to the environment file:

SIGNING_DATABASE_BOOTSTRAP_DIR=/srv/signing/database-bootstrap
SIGNING_DATABASE_INIT_SQL=/absolute/path/to/deploy/docker/roles.sql
SIGNING_DATABASE_NAME=signing

Prepare three distinct owner-protected files: database-bootstrap/admin-password, database-passwords/owner-password and database-passwords/runtime-password. Also create the volume signing_database-data with the same ownership and mode. roles.sql creates the non-superuser roles signing_owner and signing_runtime on an empty cluster only. Changing the files later does not change existing passwords.

Terminal window
docker compose -p signing --env-file /srv/signing/deployment.env \
-f deploy/docker/compose.yaml -f deploy/docker/compose.postgresql.yaml up -d --wait

The default publish address is loopback. With direct HTTPS the server uses the supplied PKCS#12 TLS identity. In private proxy mode it binds an explicit private container address, omits local TLS and relies on a proxy that terminates TLS for the configured public origin. Forwarded headers never change the public origin.

The image health check runs health-check /etc/signing/server.json. It probes only the local listener and reports healthy when the application is ready, which is stricter than a database health check.

Use docker compose stop signing for a graceful stop with a 15-second grace period. To upgrade, change the image and run up -d --wait signing. Configuration, keys and named volumes are retained.