Canary deployments
:::info No canary controller in 0.2.0
Nexo 0.2.0 (Private Preview) does not include a pipeline-level canary controller,
per-request traffic splitter, nexo canary CLI, automatic promotion, or automatic
rollback. Router-phase traffic-split execution is not active in 0.2.0. Use the options
below to stage pipeline changes safely today.
:::
Rolling out a pipeline change gradually reduces blast radius. In 0.2.0, Shadow Traffic and Traffic Split are not supported live-workload rollout controls. The standard deploy/verify/rollback flow is the current safe-rollout mechanism.
Staged rollout today
Future option — Shadow Traffic
The Shadow Traffic component mirrors a configurable fraction of live requests to a secondary target without affecting the primary response path. The client never receives the shadow response.
:::danger Do not mirror current production traffic
The current NexoShadow CRD/runtime contract is ineffective, and the runtime target path
uses plaintext TCP without target TLS. Do not use it with live, credential-bearing, or
sensitive workloads. A future evaluation must use synthetic non-sensitive data over an
independently authenticated encrypted tunnel and must verify the exact release contract.
:::
Future option — Traffic Split
The Traffic Split component distributes traffic across
weighted upstream targets via the NexoTrafficSplit CRD (Router phase). It is reconciled
by the Operator, but router-phase execution is not active in 0.2.0 — configure it now
so it is ready when router-phase support is enabled in a future release.
See the Traffic Split component page and the NexoTrafficSplit CRD for field reference.
Current option — Standard deploy/verify/rollback (all tiers)
For any pipeline change where shadow or split traffic is not available or appropriate, use the standard flow documented in Upgrade and rollback and verified against the Production checklist:
- Export and record the active pipeline revision (the rollback target).
- Apply the new pipeline revision through the console.
- Verify the active graph and confirm connection health.
- Monitor the Logging and Metrics components for latency regression or error-rate increase.
- If issues appear, reactivate the previous pipeline revision immediately.
This is the recommended default for all pipeline changes in 0.2.0.