search

pegasystems / pega-helm-charts

Orchestrate a Pega Platform™ deployment by using Docker, Kubernetes, and Helm to take advantage of Pega Platform Cloud Choice flexibility.
https://community.pega.com/knowledgebase/articles/cloud-choice
Apache License 2.0
124 stars 198 forks source link

Incorrect scaling behavior of pega-web workload #702

Closed johnbooth closed 5 months ago

johnbooth commented 7 months ago

Describe the bug When adjustments are made to either the global.tier[0].hpa.maxReplica or global.tier[0].hpa.minReplica settings within the Helm chart, the pega-web workload exhibits an unexpected behavior: it scales down to 1 replica and subsequently scales back up to the minReplica setting. This behavior poses a significant risk of service disruption and necessitates execution during a low-traffic maintenance window.

To Reproduce Update the Helm chart with modifications to global.tier[0].hpa.maxReplicas. Observe the pod rescheduling process for the pega-web workload.

Expected behavior The expectation is to seamlessly increase the maxReplica setting without triggering unnecessary scaling events that disrupt service. This aligns with the standard behavior observed when modifying HPA settings via the GCP Console.

Chart version 3.6.0 - No customizations

Server

MadhuriArugula commented 6 months ago

@johnbooth We have reviewed the use case and do not see any issue with scaling behaviour when HPA is modified or even patched. We do have enhancements provided in deployments HPA as part of latest chart releases https://github.com/pegasystems/pega-helm-charts/releases/tag/v3.16.1 where you can provide scaleUp/scaleDown strategies for HPA along with other supported parameters from Kubernetes.

Can you please try the feature and let us know if you still see the issue. If you further require assistance on the same, I recommend you to raise a Pega Support Request.

MadhuriArugula commented 5 months ago

Closing the issue as per previous comment.