Post-Deploy Issues

Hud watches every rollout from its first calls. When a new version changes a function, Hud compares that function, and the flows that run through it, with their normal from the week before. If they got much slower or started failing, Hud opens an issue and alerts you, with the function where the problem started and the forensics it captured.

For a verdict on a whole deployment, with the PR that most likely caused a problem and whether to roll back, see Deployment monitoring.

What Hud watches

Hud watches the functions a new version added or changed, and every endpoint, queue and custom flow that runs them. Hud checks within minutes of the rollout, then every hour. Each check compares the new version with the week before the rollout. Hud only checks services that are at least 3 days old, so there's a normal to compare with.

Post-Deploy Degradation

Flags an endpoint, queue or custom flow that got slower after a new version went out. Hud keeps checking for 24 hours, and by default opens an issue when, at a check:

Functions in the same flow that slowed down the same way are listed on the issue, so you can see where the slowdown started.

Post-Deploy Error

Flags errors a new version brought to the endpoints, queues and custom flows it changed. For an endpoint, an error is a 5xx response or a flow marked as failed. Hud keeps checking for 6 hours, and by default opens an issue when:

• A new error, one that didn't happen in the week before, reaches a 5% error rate with at least 5 errors.
• An existing error's rate goes up by more than 300%. Hud skips this when the error rate was already above 5% before the rollout.

The issue names the function that threw and includes the forensics Hud captured. Post-Deploy Error works with all three SDKs. On Node.js it needs SDK 1.6.28 or later, and on Python SDK 0.3.20 or later.

Change the defaults

These are Hud's defaults. To change them for a service or an endpoint, open the rule on the Rules page. Detection conditions decide when Hud opens an issue and alert conditions decide when it alerts you. While an issue keeps recurring, Hud alerts on it again every 3 days by default.

Tell one rollout from the next

Hud spots a new version on its own, from the functions whose code changed, so the Post-Deploy rules work without a deployment ID. Send one anyway to filter the web app and Hud MCP by version and to use Deployment monitoring. Use the commit SHA of the code you deploy, or a version that is also a git tag. See Add the SDK.

Next steps

• Choose where alerts go: Alerts.
• Fix the issue with your coding agent: Prompt to Fix.
• Let Hud find the root cause and open a fix PR: Auto Remediation.