iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
To make an Argo Rollouts analysis control deployment progression, define the checks in an AnalysisTemplate, let an AnalysisRun execute them, and reference that template at the rollout stage where waiting matters. For a canary, use an inline analysis step to block the next step; for blue-green, use pre-promotion analysis to hold the active-Service switch. Post-promotion analysis checks after that switch and, on failure or error, the documented behavior is to abort and switch traffic back to the previous stable ReplicaSet.
What makes an analysis a rollout gate?
An AnalysisTemplate describes measurements, how often to collect them, and the conditions that classify results as successful or failed. It is a definition, not a check that has already run. An AnalysisRun instantiates and executes the analysis; its outcome can allow the rollout to continue, abort it, or leave it paused or inconclusive. The rollout strategy must reference the template at a point that controls progression for the result to act as a gate. See the Argo Rollouts architecture documentation and analysis documentation.
Analysis is optional. A rollout can instead rely on manual pauses and promotion or other checks, including jobs and webhooks. Choose analysis when an automated measurement should inform a rollout decision.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsChoose canary or blue-green based on how you want to expose the change
Both strategies keep old and new versions involved during a rollout, but they expose the new version differently. Argo Rollouts describes blue-green as generally easier to start with; canary’s staged exposure is more configurable, but the application must tolerate old and new versions running concurrently, and fine-grained traffic splits often need traffic-management infrastructure.
#1 Best Overall
| Decision | Canary | Blue-green |
|---|---|---|
| How production exposure changes | Increase exposure to a subset of users or traffic in defined stages. | Keep a preview version alongside the active version, then switch Service selection. |
| Traffic control | Desired weights and pauses can define stages; meaningful traffic weighting often depends on a supported ingress or service-mesh integration. | Uses active and preview Services for the switch; a traffic manager is not inherently required. |
| Where analysis can gate | Inline analysis blocks at a step; background analysis runs while steps progress. | Pre-promotion analysis can hold the switch; post-promotion analysis runs after it. |
| Failure behavior | Failed analysis can abort the canary; the resulting traffic behavior depends on configuration. | Failed pre-promotion analysis prevents the switch. Failed or errored post-promotion analysis aborts and switches back to the previous stable ReplicaSet. |
The project’s strategy concepts explain the distinction. For canary traffic splitting, check the current canary documentation against the routing integration and cluster setup you actually use.
Define a check whose result means something
A passing query is only as useful as its signal. Pick a measurement that represents the behavior you want to protect during the rollout, and confirm that it produces the value and shape expected by the condition. The official Prometheus example accesses the first returned value as result[0], checks it against 0.95, samples at five-minute intervals, and sets a failure limit of three. Those are example settings in the official analysis documentation, not a recommended SLO or universal default.
Rank #2
Before relying on a metric as a gate, check that the query is representative, returns enough data during the rollout window, and has semantics that match the decision. An empty result, an unexpected result shape, or a metric unrelated to the new version can make a syntactically valid condition a poor safety check. Set success and failure criteria deliberately, and understand how the analysis handles measurements that do not meet either condition.
Free tools Windows power users keep installed
One-click scans. No signup required.
Gate a canary at the step where you need a decision
A canary strategy is a sequence of user-defined steps. setWeight specifies the desired traffic share, while pause waits for a duration or requires manual promotion. A weight does not guarantee a corresponding real traffic percentage in every setup: confirm that the configured routing integration supports the split. The official documentation notes that there is no single agreed standard for canary deployments, so the rollout steps define the progression for the application.
Inline analysis: wait before advancing
Put an analysis step after the exposure stage you want to evaluate. The controller waits for that AnalysisRun to finish before advancing: success lets the rollout proceed, while failure aborts it. This is the clearest choice when the next weight should not be set until a measured decision is available.
strategy:
canary:
steps:
- setWeight: 10
- pause:
duration: 5m
- analysis:
templates:
- templateName: success-rate
- setWeight: 25This is a strategy fragment, not a complete Rollout manifest. The referenced template must exist, and the example weights and pause duration are illustrative choices to adapt to the application and routing setup.
Rank #4
Background analysis: observe while steps continue
Background analysis runs while the canary progresses through steps rather than blocking at a dedicated inline step. Use it when ongoing monitoring is desired, but do not treat it as equivalent to a step that must complete before the next exposure increase. See the analysis strategy documentation for the supported configuration and behavior.
Gate a blue-green switch before or validate it after
Pre-promotion analysis: hold the active-Service switch
Use prePromotionAnalysis when the new version should be checked before production traffic moves to it. The rollout waits for the analysis before switching the active Service selector. A failed check therefore prevents promotion rather than discovering the problem after the switch.
strategy:
blueGreen:
activeService: app-active
previewService: app-preview
prePromotionAnalysis:
templates:
- templateName: success-rateThis fragment shows where to attach the check; it does not define the Rollout’s full pod template or the AnalysisTemplate itself.
Post-promotion analysis: check after the switch
Use postPromotionAnalysis when the check should run after the active Service has switched to the new version. On failure or error, the documented behavior is to abort and switch traffic back to the previous stable ReplicaSet. This is a post-switch validation with rollback behavior, not a gate that prevents the initial switch.
For both blue-green hooks, use the current analysis documentation to confirm the exact fields and lifecycle behavior for the version you deploy.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Quick Recap
Check the rollout wiring before relying on the gate
- Confirm the signal. Verify the provider query’s output shape and that its returned measurement reflects the version or traffic you intend to evaluate.
- Set the result rules. Define what counts as success, failure, and an insufficient or inconclusive result; choose cadence and failure handling to suit the rollout window.
- Attach the template at the decision point. Use an inline canary step or blue-green pre-promotion hook if progression must wait. Choose background or post-promotion analysis only when their timing matches the intended decision.
- Validate exposure and recovery. Confirm canary routing weights are backed by the configured integration, and understand the abort or switch-back behavior applicable to the chosen strategy.
- Check version-specific details. Strategy fields, provider behavior, and integration support can change. Consult the installation documentation and the project’s repository alongside the stable feature docs for the version and cluster in use.
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.
