In Argo Rollouts, an AnalysisTemplate defines what to measure, an AnalysisRun executes those checks, and a rollout strategy determines whether their result can stop or advance deployment. Put an analysis step in a canary when progression should wait at a particular stage; use prePromotionAnalysis in blue-green when the active Service must not switch before checks pass. Analysis documentation
How canary and blue-green expose a release
Argo Rollouts supports both strategies, but they control exposure differently. A canary introduces the new version to a subset of traffic or users and can increase exposure through defined stages. Blue-green runs the new version alongside the old one while production traffic remains on the old version, then switches the active Service to the new version. Argo Rollouts concepts
As an Amazon Associate I earn from qualifying purchases.
| Decision | Canary | Blue-green |
|---|---|---|
| How exposure changes | Increase the new version’s exposure in stages. | Prepare a preview version, then switch the active Service. |
| Traffic infrastructure | Fine-grained traffic weighting often relies on a supported ingress or service-mesh integration. Confirm the integration and its behavior for your cluster. | Uses active and preview Services; a traffic manager is not inherently required, though the application’s architecture still matters. |
| Version coexistence | The application must tolerate old and new versions running at the same time. | The versions coexist during preparation, while production initially stays on the old version. |
| Analysis timing | Run analysis alongside steps or as a blocking step. | Run analysis before promotion or after the active Service switches. |
The project documentation describes blue-green as generally easier to start with; a canary is a better fit when staged exposure and traffic control are important and the application can support simultaneous versions. Strategy concepts
Recommended Free Tools
What makes an AnalysisTemplate a real gate
An AnalysisTemplate is a reusable definition of metrics, measurement cadence, and criteria for successful or failed results. It does not run merely because it exists. An AnalysisRun instantiates the template and performs the checks. For those results to control delivery, the Rollout must reference the analysis at a strategy point whose behavior blocks or changes progression. Argo Rollouts architecture · Analysis features
#1 Best Overall
- Template: describes what to query, how often to measure, and how to interpret results.
- AnalysisRun: executes the defined checks and reports their outcome.
- Rollout reference: connects that outcome to a deployment decision, such as waiting, advancing, or aborting.
Analysis is optional. Rollouts can also be manually paused or promoted, and the architecture supports other checks such as jobs and webhooks. Architecture overview
Where analysis belongs in a canary
Canary steps let you define the sequence of exposure and pauses. A setWeight step specifies a desired traffic share, while a pause can wait for a duration or require manual promotion. The weight should not be assumed to represent a precise traffic percentage unless the deployment has an appropriate traffic-routing integration configured. Canary strategy
Rank #2
Use an inline analysis step to block progression
Place an analysis step at the point where a measured decision is required. The controller waits for that AnalysisRun to complete before continuing; a failed canary analysis can abort the rollout. This fragment shows the ordering, not a complete Rollout manifest:
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallstrategy:
canary:
steps:
- setWeight: 20
- pause:
duration: 5m
- analysis:
templates:
- templateName: success-rate
- setWeight: 50
- pause: {}The weights and pause duration above are illustrative sequencing values, not recommended defaults. The appropriate exposure increments and observation time depend on the service and its signals.
Use background analysis to monitor while steps continue
Background analysis runs while the canary progresses through its steps; it does not provide the same step-by-step wait point as inline analysis. Choose it when monitoring should span rollout activity rather than hold advancement at one particular step. Canary analysis timing
Where analysis belongs in blue-green
Pre-promotion: gate the Service switch
Use prePromotionAnalysis when the new ReplicaSet should be checked before the active Service selector changes. The analysis can hold promotion until it succeeds, keeping production traffic on the previous stable version during the check. The following fragment shows the strategy reference only:
Rank #4
strategy:
blueGreen:
prePromotionAnalysis:
templates:
- templateName: success-ratePost-promotion: validate after switching
Use postPromotionAnalysis when the check is intended to run after the active Service has switched. Argo Rollouts documents that a post-promotion analysis failure or error aborts the rollout and switches traffic back to the previous stable ReplicaSet. That is a different safety window from pre-promotion: the new version has already received production traffic before the check finishes. Blue-green analysis behavior
Make the metric and its decision rule trustworthy
A successful query is only useful if its signal represents the behavior the rollout is meant to protect. Check that the query returns enough valid data over the rollout window, that its output shape matches the condition, and that the metric is relevant to this service. A syntactically successful query does not by itself prove application health. The analysis provider and result criteria are part of the gate’s meaning. Architecture · Analysis configuration
- Verify result shape: the official Prometheus example evaluates the first returned result with
result[0]. Confirm your provider returns the structure your condition expects. - Define success and failure deliberately: criteria should reflect the signal’s units, direction, and acceptable behavior rather than being copied without context.
- Choose a cadence that fits the signal: measurements need to arrive often enough to inform the rollout, while accounting for the metric’s sampling and aggregation window.
- Account for missing or sparse data: decide how the check should behave before relying on it if the query can return no useful measurements.
The documentation’s Prometheus example uses a five-minute interval, a success condition of result[0] >= 0.95, and a failure limit of three. Those are example configuration values, not a universal service-level objective or recommended rollout policy. Prometheus analysis example
Choose the gate based on the risk you need to control
- Choose an inline canary analysis step when you need staged exposure and a specific point where rollout progression waits for evidence.
- Choose background canary analysis when checks should run while the rollout continues through its defined steps.
- Choose blue-green pre-promotion analysis when the active Service must remain on the old version until checks pass.
- Choose post-promotion analysis when validation must observe the switched version in production and the documented switch-back behavior on failure or error matches your needs.
Canary steps are deliberately user-defined rather than a single prescribed sequence; Argo Rollouts’ documentation notes that there is no agreed standard for a canary deployment. Canary strategy documentation The operational decision is therefore not simply “which strategy is safer,” but where you need exposure to change, what evidence can justify continuing, and whether your traffic and application architecture support that choice.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →
