[
https://issues.apache.org/jira/browse/FLINK-29465?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17610891#comment-17610891
]
Xin Hao commented on FLINK-29465:
---------------------------------
We currently set the pod resources request and limit the same base on the
resource definition.
For example, if we add resource definition in the FlinkDeployment CR
{code:java}
taskManager:
resource:
memory: 2048m
cpu: 0.5 {code}
The TM's pod definition will be
{code:java}
resources:
limits:
cpu: 500m
memory: 2Gi
requests:
cpu: 500m
memory: 2Gi {code}
According to the Flink KubernetesConfigOptions docs. (For example TM)
{code:java}
TASK_MANAGER_MEMORY_LIMIT_FACTOR =
ConfigOptions.key("kubernetes.taskmanager.memory.limit-factor").doubleType().defaultValue(1.0).withDescription("The
limit factor of memory used by task manager. The resources limit memory will
be set to memory * limit-factor."); {code}
With the limit factor, we can do this
{code:java}
taskManager:
resource:
memory: 512m
memoryLimitFactor: 4
cpu: 0.1
cpuLimitFactor: 2{code}
so that TM's pod spec should be
{code:java}
resources:
limits:
cpu: 100m
memory: 512Mi
requests:
cpu: 200m
memory: 2Gi{code}
> Supports limit factor in the resource definition
> ------------------------------------------------
>
> Key: FLINK-29465
> URL: https://issues.apache.org/jira/browse/FLINK-29465
> Project: Flink
> Issue Type: Improvement
> Components: Kubernetes Operator
> Reporter: Xin Hao
> Priority: Minor
>
> Can we add limit factor supports in the resource definition?
> {code:java}
> public class Resource {
> /** Amount of CPU allocated to the pod. */
> private Double cpu;
> private double cpuLimitFactor = 1.0;
> /** Amount of memory allocated to the pod. Example: 1024m, 1g */
> private String memory;
> private double memoryLimitFactor = 1.0;
> ...
> } {code}
> We can set the defaults as `1.0`, and update them to K8s config options
> {code:java}
> JOB_MANAGER_CPU_LIMIT_FACTOR
> JOB_MANAGER_MEMORY_LIMIT_FACTOR
> TASK_MANAGER_CPU_LIMIT_FACTOR
> TASK_MANAGER_MEMORY_LIMIT_FACTOR {code}
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
