A200161411/cloud-container-engine
1
0
Fork 0
forked from docs/cloud-container-engine
Code Pull Requests Activity
cloud-container-engine/umn/source/affinity_and_anti-affinity_scheduling/scheduling_policy_overview.rst
proposalbot a70a2c8b2e Changes to cce_umn from docs/doc-exports#418 (CCE UMN for 1.23 reuploaded -20221 
Reviewed-by: Hasko, Vladimir <vladimir.hasko@t-systems.com>
Co-authored-by: proposalbot <proposalbot@otc-service.com>
Co-committed-by: proposalbot <proposalbot@otc-service.com>
2022-11-10 18:57:33 +00:00

3.7 KiB
Raw Blame History

original_name

cce_01_0051.html

Scheduling Policy Overview

Custom Scheduling Policies

You can configure node affinity, workload affinity, and workload anti-affinity in custom scheduling policies.

  • Node Affinity <cce_01_0232>
  • Workload Affinity <cce_01_0233>
  • Workload Anti-Affinity <cce_01_0234>

Note

Custom scheduling policies depend on node labels and pod labels. You can use default labels or customize labels as required.

Simple Scheduling Policies

A simple scheduling policy allows you to configure affinity between workloads and AZs, between workloads and nodes, and between workloads.

  • Workload-AZ affinity: Multiple AZ-based scheduling policies (including affinity and anti-affinity policies) can be configured. However, scheduling is performed as long as one of the scheduling policies is met.
    • Affinity between workloads and AZs: Workload-AZ Affinity <cce_01_0228>
    • Anti-affinity between workloads and AZs: Workload-AZ Anti-Affinity <cce_01_0229>
  • Workload-node affinity: Multiple node-based scheduling policies (including affinity and anti-affinity scheduling) can be configured. However, scheduling is performed as long as one of the scheduling policies is met. For example, if a cluster contains nodes A, B, and C and two scheduling policies are set (one policy defines node A as an affinity node and the other policy defines node B as an anti-affinity node), then the workload can be scheduled to any node other than B.
    • Affinity between workloads and nodes: Workload-Node Affinity <cce_01_0225>
    • Anti-affinity between workloads and nodes: Workload-Node Anti-Affinity <cce_01_0226>
  • Workload-workload affinity: Multiple workload-based scheduling policies can be configured, but the labels in these policies must belong to the same workload.

    • Affinity between workloads: For details, see Workload-Workload Affinity <cce_01_0220>. You can deploy workloads on the same node to reduce consumption of network resources.

      Figure 1 <cce_01_0051__fig3017424713> shows an example of affinity deployment, in which all workloads are deployed on the same node.

      Figure 1 Affinity between workloads

    • Anti-affinity between workloads: For details, see Workload-Workload Anti-Affinity <cce_01_0227>. Constraining multiple instances of the same workload from being deployed on the same node reduces the impact of system breakdowns. Anti-affinity deployment is also recommended for workloads that may interfere with each other.

      Figure 2 <cce_01_0051__fig1505421971> shows an example of anti-affinity deployment, in which four workloads are deployed on four different nodes.

      Figure 2 Anti-affinity between workloads

Important

When setting workload-workload affinity and workload-node affinity, ensure that the affinity relationships do not contradict each other; otherwise, workload deployment will fail.

For example, Workload 3 will fail to be deployed when the following conditions are met:

  • Anti-affinity is configured for Workload 1 and Workload 2. Workload 1 is deployed on Node A and Workload 2 is deployed on Node B.
  • Affinity is configured between Workload 2 and Workload 3, but the target node on which Workload 3 is to be deployed is Node C or Node A.