Bug Fixes
🐞 Unquoted cluster name could cause deployment failures 🐞
Fixes a bug raised in #193 where some cluster names that helm interprets as non-strings could cause deployment failures.
🐞 Fixes typo with trn2u.48xlarge for neuron-monitor 🐞
Fixes a typo in the nodeSelector values for neuron-monitor with trn2u.48xlarge.
Enhancements
💡 Manage tolerations at a component level 💡
You can now specify tolerations for each of the individual components deployed by the helm chart.
The daemonsets (cloudwatch-agent, aws-for-fluent-bit, dcgm-exporter, neuron-monitor) deployed by the helm chart default to the root level tolerations, where they tolerate all taints.
The deployments (amazon-cloudwatch-agent-operator) specify empty tolerations at the component level to override the default at the root level.
💡 Manage affinity and nodeSelectors at a component level 💡
You can now specify the affinity and nodeSelectors for the individual components deployed by the helm chart.
💡 Manage updateStrategy at a component level for cloudwatch-agent and aws-for-fluent-biit 💡
You can now specify the updateStrategy for cloudwatch-agent and aws-for-fluent-biit deployed by the helm chart.
Component Versions
- cloudwatch-agent - v1.30055.0b1095
- aws-for-fluent-bit - v2.32.5.20250327
- dcgm-exporter - 3.3.9-3.6.1-ubuntu22.04
- neuron-monitor - v1.4.0
- adot-autoinstrumentation-java - v2.10.0
- adot-autoinstrumentation-python - v0.9.0
- adot-autoinstrumentation-dotnet - v1.7.0
- adot-autoinstrumentation-node - v0.6.0