You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Kubernetes deployments whose Neo4j Enterprise or PostgreSQL task-manager database runs outside the cluster can now be backed up and restored with the same infrahub-backup create and restore commands as any other deployment. Artifacts taken from an external database are byte-compatible with internal ones and restore by the same path, and deployments whose databases are all internal behave exactly as before. When the deployment's own pods take the database username and password from named Secret keys (env[].valueFrom.secretKeyRef), the backup reads the same keys and needs no permissions on secrets. Credentials from any other source, including envFrom, need permission to create a Secret for the run. When infrahub-backup itself runs as a pod in the Infrahub namespace, it runs the database tools from its own image and creates no Job or Secret, so it needs only the permissions of a backup of internal databases. (#172)
Fixed
On Docker, a restore now waits for the database container to be healthy again before it starts infrahub-server and task-worker, and a failed docker compose start or stop reports what compose printed instead of only its exit status. (#172)
The Kubernetes backup permissions now list list on deployments and statefulsets and get and patch on deployments/scale and statefulsets/scale. A Role that granted only get and patch on the workloads, as the documentation previously said, cannot stop and restart the Infrahub components, because kubectl scale writes to the scale resources.
When --k8s-namespace is set and the namespace cannot be checked, the error now includes what kubectl reported, such as a forbidden refusal or a missing namespace, instead of only exit status 1.
When --k8s-namespace or --project is set, the tools now act only on that deployment. Previously, if the Kubernetes namespace could not be checked, they went on to detect a Docker Compose project on the same host and could back up or restore that project instead.