Log inStart free

Example

Ephemeral Kubernetes targets (per-run namespace + database)

Guidance, not a separate integration: the CI recipes stay the same; what changes is the target pool.

About this example

Guidance, not a separate integration: the CI recipes stay the same; what changes is the target pool.

  1. Pre-create N per-run target slots with the self-hosted ephemeral PostgreSQL target that ships with the agent's deployment examples (agent installation): one namespace per slot, restricted Pod Security, default-deny NetworkPolicy, one PostgreSQL each.
  2. Register each slot's connection settings in your secret provider and list the references in the profile's lease.target_pool (for example vault://ci/k8s-slot-1 … vault://ci/k8s-slot-8). The profile is approved by a second person; a pipeline can never add a slot.
  3. Allow the CI runner (and the DataNivra agent) to reach those namespaces only.

Every request then gets one free slot; release (or TTL expiry) makes the agent tear the copy down, and the slot returns to the pool. DataNivra does not create namespaces or clone databases itself (no native clone: ADR-0016); per-lease isolation comes from the separate slots.