Migrating from legacy to split health checks

As of September 15, 2019, if you're using the legacy health checks, your application will continue to run and receive health checks but you won't be able to deploy new versions of your application.

This page explains upgrading from the legacy health checks to the split health checks.

Verifying your health check type

To verify the type of health checks your application is using, run the following command:

gcloud app describe

If your application is using split health checks, the description should include the following information:

featureSettings:
    splitHealthChecks: true

Understanding key differences

Before upgrading to split health checks, consider the following important differences between legacy and split health checks:

  • HTTP requests for split health checks are not forwarded by default. By contrast, legacy health checks forward to the /_ah/health path in your application by default.

  • Forwarded split health checks must return 200 OK when healthy and ready. Legacy health checks consider the following HTTP codes as healthy: 200, 301, 302, 303, 307, 401, 402, 403, 404, 405.

If you do not specify a liveness check path or a readiness check path, by default split health checks only confirm that the VM instance and the Docker container are running. As long as these conditions hold, the VM will continue to receive traffic and remain alive regardless of the app's internal state.

By contrast, when legacy health checks are enabled, if your app's /_ah/health path starts returning unhealthy HTTP error codes (e.g. 5XX) then legacy health checks will start failing and the VM will stop receiving traffic and be restarted.

If your app depends on the default legacy health check behavior, set the liveness check path and the readiness check path accordingly.

Converting legacy health check options

Each legacy health check option can be rewritten using split health checks as follows:

Option Maintaining same behavior in split health checks
enable_health_check If True or unset, configure liveness_check.path and readiness_check.path to a path on the app that returns 200 OK when the app is healthy.
check_interval_sec Configure liveness_check.check_interval_sec and readiness_check.check_interval_sec to the same value.
timeout_sec Configure liveness_check.timeout_sec and readiness_check.timeout_sec to the same value.
unhealthy_threshold Configure readiness_check.failure_threshold to the same value.
healthy_threshold Configure liveness_check.success_threshold and readiness_check.success_threshold to the same value.
restart_threshold Configure liveness_check.failure_threshold to the same value. Note the value of your check_interval_sec option multiplied by the failure_threshold option is the time it'll take for an unhealthy VM to be taken down.

Enabling split health checks

To migrate from legacy health checks to split health checks and avoid seeing elevated 5xx status codes, complete the following steps:

  1. Understand the important differences between legacy and split health checks.

  2. Convert legacy health check options for each version in your application.

    Alternatively, you can customize the liveness_check or readiness_check section in the app.yaml file for each version. For examples, see Liveness checks and Readiness checks.

  3. Run the following command:

    gcloud app update --split-health-checks --project [YOUR_PROJECT_ID]
  4. If you used customized settings for legacy health checks, you must remove the health_check section from your app.yaml file.

  5. Deploy a new major version of your app to start using liveness and readiness health checks.

本頁內容對您是否有任何幫助?請提供意見:

傳送您對下列選項的寶貴意見...

這個網頁
App Engine flexible environment for Go