Skip to main content

Cloud Hosting FAQ

Why are old version instances of Cloud Hosting not automatically destroyed after deploying a new version?​

After creating a new version instance, the old version instance cannot be scaled down immediately; otherwise, existing connections and traffic before the new version is successfully started will cause errors.
If multiple versions exist online and all have traffic, automatic scaling down will not occur.

Instance scale-down must meet the following two conditions:

  • No traffic for 30 minutes (only effective when the minimum number of replicas is set to 0)
  • Delayed decommissioning and scaling down (when the version has no traffic)

Note: The configured maximum number of replicas controls the pod of a single version. If multiple versions exist, then:
`Maximum number of pods = Configured maximum number of replicas × Number of versions

Forced scale-down can also be achieved by manually deleting old versions.

According to monitoring, both CPU and memory usage have not reached the set scale-out threshold. Why did Cloud Hosting automatically scale out?​

In monitoring, you can view CPU and memory usage. However, the memory usage displayed in monitoring does not include Page Cache. Data in Page Cache also occupies memory. If the usage consistently remains above the set value, scaling down will not be possible.

Understanding Page Cache:

  • In Kubernetes, the Cache in a Pod's memory usually refers to the Page Cache used by applications.
  • When applications read from or write to files, Page Cache caches file data and metadata to improve file access performance.
  • If the application in a Pod frequently reads and writes large volumes of files, Page Cache may occupy substantial memory, resulting in excessively high memory usage for the Pod.

Similar Issues​

  • Why are old version instances of Cloud Hosting not automatically destroyed after deploying a new version?
  • Under what conditions will old version instances of Cloud Hosting scale down?
  • What is the reason for Cloud Hosting's automatic scale-out?
  • What impact does Page Cache have on Cloud Hosting memory usage?
  • How to manually force scale-down of old version instances in Cloud Hosting?
  • How is the maximum replica count calculated in Cloud Hosting?
  • Why is the memory usage in Cloud Hosting monitoring inaccurate?
  • How to avoid unnecessary automatic scale-out in Cloud Hosting?
  • What is Page Cache in Cloud Hosting?
  • How to control the replica count in multi-version Cloud Hosting?