跳到主要内容
版本：v2.12

UI Server-Side Pagination

Server-Side Pagination (SSP) is a Rancher feature to provide significant performance improvements across the UI for resources with high counts, restricting the amount of resources browser fetches and stores in memory.

Note that SSP is optional, enabled by default, and it can be disabled via the feature flag ui-sql-cache.

Disk Space​

信息

It is crucial that you review the available disk space on your nodes and plan accordingly before upgrading to Rancher v2.12.0 and later to avoid potential disk pressure and pod eviction issues.

The SSP relies on a caching mechanism that introduces a new requirement for ephemeral disk space on your cluster nodes. This cache, an internal SQLite database, is stored within the container's file system. This affects the nodes running the Rancher server pods (rancher in the cattle-system namespace on the local cluster) and the nodes running the Rancher agent pods (cattle-cluster-agent in the cattle-system namespace on all downstream clusters).

The amount of disk space required is dynamic and depends on the quantity and size of Kubernetes resources visualized in the UI. As a guideline, the cache may consume approximately twice the size of the raw Kubernetes objects it stores.

For example, internal tests showed that caching 5000 ConfigMaps, totaling 50 MB, consumed 81 MB of disk space. For a conservative, high-level estimate, you can plan for the available disk space on each relevant node to be at least twice the size of your etcd snapshot. For most production environments, ensuring a few extra gigabytes of storage are available on the relevant nodes is a safe starting point.

Please note this space counts against ephemeral storage requests and limits you might have set for your Rancher container via the resource value in the Helm chart. Make sure those settings provide for abundant available space.

If you see the error database or disk is full (13) in the pod logs, this is a symptom that more space needs to be allocated.

SQLite-backed caching persists copies of any cached Kubernetes objects to disk. See Encrypting SQLite-backed Caching if this is a security concern.

Enabling Server-Side Pagination​

  1. In the upper left corner, click ☰ > Global Settings > Feature Flags.
  2. Find ui-sql-cache and select ⋮ > Activate > Activate.
  3. Wait for Rancher to restart. This also restarts agents on all downstream clusters.
  4. Reload the page with the browser button (or the equivalent keyboard combination, typically CTRL + R on Windows and Linux, and ⌘ + R on macOS).

Disabling Server-Side Pagination​

  1. In the upper left corner, click ☰ > Global Settings > Feature Flags.
  2. Find ui-sql-cache and select ⋮ > Deactivate > Deactivate.
  3. Wait for Rancher to restart. This also restarts agents on all downstream clusters.
  4. Reload the page with the browser button (or the equivalent keyboard combination, typically CTRL + R on Windows and Linux, and ⌘ + R on macOS).

Encrypting SQLite-backed Caches​

UI server-side pagination persists copies of any cached Kubernetes objects to disk. If you're concerned about the safety of this data, you can encrypt all objects before they are persisted to disk, by setting the environment variable CATTLE_ENCRYPT_CACHE_ALL to true in rancher pods in the upstream cluster and cattle-cluster-agent pods in the downstream clusters.

Secrets and security Tokens are always encrypted regardless of the above setting.

Known Limitations of UI Server-Side Pagination​

This release improves the performance of most pages used to view, create or edit resources within the local or downstream clusters i.e. the Cluster Explorer view. However, RBAC related resources and areas outside the Cluster Explorer are not yet covered by this feature.

Additionally, the following limitations are present when the feature is enabled. These mainly revolve around different sort or filter behaviors in affected lists:

  • Resources in lists are automatically updated, however, not instantaneously.
  • All lists that utilize Server-Side Pagination:
    • State column sort and filter features work on the resources metadata.state.name field instead of one deduced locally by the UI.
    • Updates are shown every 5 seconds, rather than instantly.
  • Cluster Explorer:
    • Cluster group --> Nodes page
      • The following columns are not sortable or filterable: Roles, External/Internal IP, CPU, RAM (logic to determine their value is calculated in the browser)
    • Workloads list:
      • The Workloads list, which showed multiple different resource types has been removed.
        • Server-Side Pagination of multiple resources is not currently possible.
    • Workloads group --> All lists
    • Workloads group / Job List
    • Workloads group / Pod List
      • Images is not sortable (sorting on an array).
    • Service Discovery group / Ingresses
      • Default is not sortable/filterable (logic to determine their value is calculated in the browser).
    • Storage group / ConfigMaps
      • Data is not sortable/filterable (logic to determine their value is calculated in the browser).
    • Storage group / Secrets
      • Data is not sortable/filterable (logic to determine their value is calculated in the browser).