Logstash chart should support pod topologySpreadConstraints

[jgregmac]

Feature

The Logstash chart should support specification of topologySpreadConstraints on the pods created by the statefulsets template.

I will file a PR that supports this feature.

Justification

In many cases, we deploy Logstash releases that create a number of pods that exceed the number of nodes in the target cluster. This is the case when using Logstash to process large workloads from a message queue. In the interest of not overloading any one node of the cluster, we want to spread out the scheduling of these pods evenly across all worker nodes in the Kubernetes cluster.

The existing affinity: podAntiAffinity rules are not helpful under these circumstances. podAntiAffinity effectively will spread out the pods only as long as the number of pods does not exceed the count of worker nodes in the cluster. After that, podAntiAffinity actually blocks scaling of the statefulSet.

topologySpreadConstraints , by contrast, allow us to more flexibly inform the scheduler how we want workloads distributed on the cluster, and allow us to specify the whenUnsatisfiable: ScheduleAnyway option to signal to the scheduler that the workload should be deployed even if the ideal pod distribution cannot be achieved.

References

• https://kubernetes.io/docs/concepts/workloads/pods/pod-topology-spread-constraints/

[jgregmac]

A PR implementing this feature has been submitted. I have created a few releases in my own cluster off of this code branch, and it seems to be working effectively.
See: #1514

[botelastic]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[botelastic]

This issue has been automatically closed because it has not had recent activity since being marked as stale.