Add pipeline for system metrics

[sky333999]

Description of changes

• Adds a new OTEL pipeline and receiver that collects JVM heap metrics and host system metrics on EC2 Linux hosts, publishing them to CloudWatch under the CWAgent/System namespace.
• Pipeline: systemmetrics → ec2tagger → batch(15m) → awscloudwatch
• Pipeline only setup in any of the following cases:
  • explicitly enabled via agent.system_metrics_enabled config field
  • explicitly enabled via SYSTEM_METRICS_ENABLED
  • not running in containerized envs and is running on a recognized instance
• All metrics published with fixed schema - InstanceId dimension and No dimensions (i.e. aggregated to the account level)

Published metrics (18 total)

Metric	Source	Unit	Scope
heap_max_bytes	jvm_heap_committed_bytes from socket	Bytes	Per-JVM
heap_committed_bytes	jvm_heap_committed_bytes from socket	Bytes	Per-JVM
heap_after_gc_bytes	jvm_heap_after_gc_bytes from socket	Bytes	Per-JVM
heap_free_after_gc_bytes	max − afterGC	Bytes	Per-JVM
aggregate_heap_after_gc_utilized	SUM(afterGC) / SUM(max) × 100	Percent	Per-Box
aggregate_heap_max_bytes	SUM(max)	Bytes	Per-Box
aggregate_heap_free_after_gc_bytes	SUM(max) − SUM(afterGC)	Bytes	Per-Box
aggregate_jvm_count	Count of discovered JVMs	None	Per-Box
mem_total	gopsutil /proc/meminfo MemTotal	Bytes	Per-Box
mem_available	gopsutil /proc/meminfo MemAvailable	Bytes	Per-Box
mem_cached	gopsutil /proc/meminfo Cached	Bytes	Per-Box
mem_active	gopsutil /proc/meminfo Active	Bytes	Per-Box
cpu_time_iowait	gopsutil /proc/stat (delta-based, 8-state denominator)	Percent	Per-Box
aggregate_disk_used	gopsutil disk.Partitions/Usage (aggregated, /dev/ filter)	Bytes	Per-Box
aggregate_disk_free	gopsutil disk.Partitions/Usage (aggregated, /dev/ filter)	Bytes	Per-Box
aggregate_bw_in_allowance_exceeded	safchain/ethtool (delta, summed across interfaces, skip lo/veth)	None	Per-Box
aggregate_bw_out_allowance_exceeded	safchain/ethtool (delta, summed across interfaces, skip lo/veth)	None	Per-Box
aggregate_pps_allowance_exceeded	safchain/ethtool (delta, summed across interfaces, skip lo/veth)	None	Per-Box

CloudWatch exporter changes

Added optional max_retry_count, backoff_retry_base, and max_concurrent_publishers config fields to the CloudWatch exporter. Zero values preserve existing defaults. The system metrics pipeline uses conservative settings: 2 retries, 1-minute backoff base, single concurrent worker — minimizing PutMetricData API pressure.

License

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

Tests

• Unit tests added/updated
• Integ tests added: Add system metrics integ tests (enabled/disabled) amazon-cloudwatch-agent-test#661
• GH run: https://github.com/aws/amazon-cloudwatch-agent/actions/runs/23064139336

Requirements

Before commiting your code, please do the following steps.

1. Run make fmt and make fmt-sh
2. Run make lint

---

Integration Tests

To run integration tests against this PR, add the ready for testing label.

[chadpatel]

I reviewed agentically by comparing to the other baseline, looks good.