Backup
GES keeps documentation catalogs in GitLab, so GitLab backs them up. The database, settings, portal data and AI service data stay on the GES server — back them up separately. Make backups regularly and always before an upgrade.
What is stored in GitLab
You don't need to copy this data separately — it is included in your GitLab backup:
Catalogs — repositories with articles.
GES settings with
ENTERPRISE_STORAGE_TYPE=gitlab— space, user and access settings files in theGIT_PROJECT_PATHrepository.
Make sure GitLab backup is set up and includes these repositories.
What is stored on the server
In docker-compose.yaml, service data is kept in Docker volumes and in a folder next to the file:
enterprise-postgres-data— thepostgresservice volume. It holds the GES database: space metrics, training module results, access tokens.enterprise-config— theenterpriseservice volume, folder/app/config. WithENTERPRISE_STORAGE_TYPE=local, it holds all GES settings../docportal— the folder next todocker-compose.yaml, mounted to thedocportalservice. It holds the documentation portal data.caddy_dataandcaddy_config— thecaddyservice volumes. They hold TLS certificates and the Caddy configuration.ges-ai-qdrant-storage— theqdrantservice volume, the vector database of AI features. Present if theaiprofile is on.ges-ai-api-storage— theges-aiservice volume, folder/app/storage. Present if theaiprofile is on.
Besides service data, save the files in the installation folder:
.env— all variables: license key, tokens,COOKIE_SECRET,ENTERPRISE_SERVICE_ENCRYPTION_KEY, database password. Keep the copy in a secure place.docker-compose.yaml— it contains your domains in the Caddy configuration and the variables you added to theenvironmentsections.custom-certs.pem, if you mounted your own root certificates.
Back up with Docker Compose
Run all commands in the folder with docker-compose.yaml.
Save the database. You can run the command while GES is running:
docker compose exec -T postgres sh -c 'pg_dump -U "$POSTGRES_USER" -d "$POSTGRES_DB"' > ges-db.sqlStop the services so that volume data does not change while copying:
docker compose --profile ai stopFind the exact volume names. Docker Compose prefixes them with the project name — by default, the name of the folder with
docker-compose.yaml:docker volume lsArchive each volume. Example for the
enterprise-configvolume in thegesproject:docker run --rm -v ges_enterprise-config:/data -v "$PWD":/backup postgres:18-alpine \ tar czf /backup/enterprise-config.tar.gz -C /data .The example uses the
postgres:18-alpineimage because it is already on the server. Any image withtarworks. Archivecaddy_data,caddy_configand, if AI features are on,ges-ai-qdrant-storageandges-ai-api-storagethe same way.Archive the portal folder and the installation files:
tar czf ges-files.tar.gz docportal .env docker-compose.yamlStart the services:
docker compose --profile ai up -d
Without AI features, run the commands without --profile ai. Move the archives and ges-db.sql to separate storage.
Restore with Docker Compose
Extract
ges-files.tar.gzinto the installation folder.Create the volumes and containers without starting them:
docker compose --profile ai createExtract the archives into the volumes. Example for
enterprise-config:docker run --rm -v ges_enterprise-config:/data -v "$PWD":/backup postgres:18-alpine \ tar xzf /backup/enterprise-config.tar.gz -C /dataStart only the database and load the copy into it. The
enterprise-postgres-datavolume must be empty — thenpostgrescreates an empty database from the.envvariables:docker compose up -d postgres docker compose exec -T postgres sh -c 'psql -U "$POSTGRES_USER" -d "$POSTGRES_DB"' < ges-db.sqlStart the other services:
docker compose --profile ai up -dCheck the service status as described in Service health checks.
Install with Helm
In Kubernetes, service data is kept in PersistentVolumeClaims created by the chart. The data is the same as in Docker Compose: GES settings, portal data, AI service data.
Back up the PVCs with your cluster tools, for example volume snapshots.
If PostgreSQL runs separately, back it up with your PostgreSQL server tools, for example
pg_dump.Save
values.yamland the Kubernetes secrets you created as described in Create Secrets.