Skip to main content
Version: 3.7.x

Release Notes

3.7.1​

Bug Fixes​

  • CLOUD-3791: Updated the recommended Amazon RDS for PostgreSQL engine versions used by the create-and-manage-database-server Jenkins job. The postgres-rds option now uses PostgreSQL 14.23, and postgres-rds-16 now uses PostgreSQL 16.10. The job also fails fast with a clear validation error when a requested RDS engine version is unavailable.
  • CLOUD-3753: Ensure clean ActiveMQ broker shutdown during pod termination and rollouts, even with ActiveMQ containers that don't properly pass signals to the broker Java process. Unclean shutdowns could log "Recovering pageFile free list due to prior unclean shutdown" and risk KahaDB corruption. ActiveMQ deployments now always find the broker process and send it SIGTERM via a Kubernetes preStop hook. The pod termination grace period is also increased from the Kubernetes default of 30s to 180s so the broker can checkpoint and flush before SIGKILL. After upgrading, redeploy existing ActiveMQ containers (for example, with the create-or-delete-activemq-containerjob) so they pick up the fix. An image rebuild is not required.
  • CLOUD-3821: Fixed an issue where ActiveMQ deployments failed to become Ready when CloudWatch Application Signals Java auto-instrumentation was enabled via the instrumentation.opentelemetry.io/inject-java: "true" pod annotation. ActiveMQ liveness and readiness probe commands now unset JAVA_TOOL_OPTIONS so the health-check subprocess does not load the OpenTelemetry javaagent and probes complete within the configured timeout.
  • CLOUD-3827: Improved on the CLOUD-3758 fix by creating and using a dedicated priority class for the modsecurity-spoa DaemonSet instead of depending on the shared ep-high class. Also elevated the scheduling priority for the HAProxy Ingress and CloudWatch logging agents by assigning them the ep-high priority class.
  • CLOUD-3841: Fixed an issue where ActiveMQ containers failed to start because JDK_JAVA_OPTIONS was set to the literal string "null". This happened when ops-spec.json (from Self-Managed Commerce or the CloudOps for Kubernetes fall-back file) lacked .activemq.jvmRuntimeArguments; eplib.parseOpsSpec passed jq's literal null into the ActiveMQ image build. Parsing now treats a missing or null value as an empty string, and the fall-back ops-spec.json defines .activemq.jvmRuntimeArguments. After upgrading, rebuild ActiveMQ images and redeploy existing ActiveMQ containers so they pick up the fix. See ops-spec.json.

New Features​

  • CLOUD-3843: Exposed the ActiveMQ container memory and CPU values as user-configurable Terraform variables, and updated the default memory requests and limits to give the ActiveMQ process more native memory. See Configure ActiveMQ Container.

3.7.0​

New Features​

  • CLOUD-3025: Added optional PodDisruptionBudgets (PDBs) for Cortex, Search Slave, and Integration to reduce outages during nodegroup rebuilds and other drain operations. PDBs are enabled by default for Cortex and Search Slave, and are opt-in for Integration. The rebuild-nodegroups.sh drain script is also hardened with per-node pacing and a 30-minute timeout. For full details, see Auto Scaling and Replicas.
  • CLOUD-3275: Added the generate-support-report Jenkins job, which produces a report.zip artifact containing Terraform workspace state, ECR repository information, and Kubernetes diagnostics to assist with infrastructure troubleshooting and support.
  • CLOUD-3333: Changed the default AWS database instance type from db.r5.xlarge to the more modern and cost-effective db.r6g.xlarge. The default value of the parameter named dbInstanceClass, in Jenkins job create-and-manage-database-server, is updated to db.r6g.xlarge. New database instances created with this job will default to the new instance type unless the parameter is changed. You can update existing instances to use the new type by rerunning the job with the updated dbInstanceClass parameter; changes will take effect during the configured AWS maintenance window.
  • CLOUD-3557: Added support for deploying a bastion instance in a private subnet, where previously bastions were deployed only in public subnets. This is controlled by the new subnetType parameter in the create-and-manage-bastion-instance Jenkins job.
  • CLOUD-3580: Updated the Jenkins server image version to 2.541.2-lts-jdk21 and the Jenkins helm chart version to 5.8.142.
  • CLOUD-3622: Replaced the legacy New Relic agent deployment with the officially supported nri-bundle Helm chart, and updated the default New Relic version to 6.0.40. The agent continues to be controlled by the existing TF_VAR_enable_new_relic_agent and TF_VAR_enable_new_relic_k8s_data parameters. For more information about the New Relic Infrastructure (NRI) Helm chart deployment, see nri-bundle.
  • CLOUD-3628: Added a file named AGENTS.md to the root of the cloud-ops-kubernetes Git project to help coding agents better understand the content, purpose and structure of the project. For more information about AGENTS.md files, see AGENTS.md.
  • CLOUD-3641: Updated CloudOps for Kubernetes to provision and operate Amazon EKS 1.35. The update includes compatible Kubernetes add-ons and bootstrap tools: eksctl; kubectl; Helm; Cluster Autoscaler; kube-proxy; CoreDNS; Metrics Server; Amazon Elastic File System Container Storage Interface Driver; kube-state-metrics; and the EKS pause image. For upgrade details, see Update to Version 3.7.
  • CLOUD-3684: Added a new page about granting additional AWS users and roles access to view the cluster in the Amazon EKS console and to run kubectl commands. See Grant Users and Roles Access to EKS.
  • CLOUD-3688: Added an Eclipse P2 caching proxy service to improve Self-Managed Commerce build stability by drastically reducing external Eclipse downloads across builds. It is enabled by default but can be disabled if desired. For more information, see Eclipse P2 Caching Proxy.
  • CLOUD-3694: Added AWS Secrets Manager support for Jenkins credentials. This helps preserve customer-managed credentials across CloudOps for Kubernetes upgrades, including credentials added through the Jenkins web interface. For upgrade details, see Update to Version 3.7. For credential management procedures, see Manage Jenkins Credentials.
  • CLOUD-3716: Updated the Terraform version from 0.14.11 to 1.8.2. Updated the Terraform provider version constraints used by the bundled modules to current supported versions, including AWS ~> 5.0, Kubernetes ~> 2.29, and Helm ~> 2.12, and included compatibility fixes required by the newer providers. No additional upgrade tasks are required beyond the normal CloudOps for Kubernetes upgrade process. If your team runs Terraform directly against the CloudOps for Kubernetes code, ensure that you use the Terraform version required by the checked out release.
  • CLOUD-3724: Added support for SAML integration with Jenkins, so teams can use Single Sign On (SSO) to control Jenkins authentication and authorization. This has several advantages over using Jenkins native login features, including easier on-boarding and off-boarding of team members, and ensures that user credentials are not lost when Jenkins is reconfigured from infrastructure code. For more information, see Jenkins SAML SSO.
  • CLOUD-3759: Added diagnostic commands to the bootstrap/scripts/rebuild-nodegroups.sh script, to provide additional information about the node and pod states during EKS node group rebuild maintenance.

Bug Fixes​

  • CLOUD-3206: Updated the bootstrap container's entrypoint.sh script to always clear the local Terraform state that manages the backend S3 bucket. This change is to avoid error message like kubernetes-bootstrap | Error: error deleting S3 Bucket (bucketname): BucketNotEmpty: The bucket you tried to delete is not empty that might occur when running docker-compose up.
  • CLOUD-3595: Added validation to the commerce-branch-validation Jenkins job so missing branch parameters fail fast with clear error messages.
  • CLOUD-3639: Switched the HAProxy Ingress implementation from a Kubernetes deployment to a daemonset, to improve availability and uptime when HAProxy Ingress is restarted.
  • CLOUD-3623: Updated the multi-purpose-commerce-tool Jenkins job to use a dedicated activeMqResourcingProfile parameter for ActiveMQ deployments, separate from Commerce stack resourcing. Added early validation to ensure the selected ActiveMQ profile exists in terraform/activemq/env-file, with clear error messaging when the profile is missing.
  • CLOUD-3661: Removed legacy ActiveMQ HA deployment behavior and updated ActiveMQ deployments to use the Kubernetes Recreate strategy during updates. This prevents a second broker pod from being started during rollout, which avoids KahaDB lock contention.
  • CLOUD-3690: Changed the HAProxy-Ingress default timeouts from fifty seconds (50s) to seventy seconds (70s), to address "504 (Gateway Timeout)" issues reported in the Commerce Manager application.
  • CLOUD-3695: Removed American National Standards Institute (ANSI) colour codes from bootstrap console output. The ANSI colour codes were adding extra printed characters to console output when redirected to files.
  • CLOUD-3736: Corrected an issue in the database Terraform modules where, in a specific scenario, Terraform would mark the existing in-use Key Management Service (KMS) key for deletion. This would only occur if the user reran the create-and-manage-database-server job and manually specified the existing Terraform-managed KMS key in the encryptionKey parameter.
  • CLOUD-3750: Fixed issues in the bootstrap/scripts/remove-nodegroups.sh script that intermittently caused node group CloudFormation stack deletion to fail due to security group associations.
  • CLOUD-3758: Fixed an issue where all pods in the modsecurity-spoa DaemonSet might not start up correctly after node group changes by assigning the DaemonSet an elevated PriorityClass.
  • CLOUD-3775: Removed the unnecessary ReservedCodeCacheSize MAVEN_OPTS setting in the Maven Java build jobs, for compatibility with Self-Managed Commerce release 8.8.x.
  • CLOUD-3776: Specified all Jenkins plugins to be installed, including the specific versions, to ensure that compatible plugins are always selected and installed.
  • SUP-6337: Updated the Jenkins controller Java memory configuration so the maximum heap is set at 60% of available memory instead of 87.5%. This makes more memory available to non-heap components to avoid Kubernetes OOMKilled issues.

Deprecations & Removals​

  • CLOUD-3656: Removed code that was used to create and manage some remaining Account Management API resources. Support for Account Management API ended with CloudOps for Kubernetes release 3.1.
  • CLOUD-3611: Removed the Alert Logic daemonset from CloudOps for Kubernetes. Alert Logic support was deprecated in CloudOps for Kubernetes release 2.14.

See Deprecations and Removals.

Upgrade Instructions​

For upgrade instructions, see Upgrading CloudOps for Kubernetes.