How do I set reasonable resource limits to prevent loss of logs?

[txjjjjj]

Because we only use agent to collect eks logs, we don't need to collect metrics and trace.

1. Is there any documentation on stress testing ? we need to set more reasonable resource limits.
2. How to avoid not losing logs, is there any best practice. Because the same cluster may contains node group instances resource configuration gap is very large.
3. We have some arm nodes in our eks cluster, is it recommended to use otel agent?

btw, I noticed some default configurations

    exporters:
      otlphttp:
        endpoint: http://${LOGS_METADATA_SVC}.${NAMESPACE}.svc.cluster.local.:4318
        sending_queue:
          queue_size: 10

    exporters:
      sumologic/containers:
        endpoint: ${SUMO_ENDPOINT_DEFAULT_LOGS_SOURCE}
        json_logs:
          add_timestamp: false
        log_format: json
        sending_queue:
          enabled: true
          num_consumers: 10
          queue_size: 10000
          storage: file_storage

daemonset
        resources:
          limits:
            cpu: 1000m
            memory: 1Gi
          requests:
            cpu: 100m
            memory: 32Mi


statefulset
        resources:
          limits:
            cpu: 1000m
            memory: 1Gi
          requests:
            cpu: 500m
            memory: 768Mi