dest_workload_name unknown

[letian0805]

What happened?

Description

When accessed via IP, the monitoring data of the TCP connection shows dest_workload_name="(unknown)".

Steps to Reproduce

1. The service in the container accesses an IP address.
2. View the tcp_bytes metric.

Expected Result

dest_workload_name should be a domain name or IP address, not unknown.

Actual Result

(unknown)

eBPF Collector version

v0.10.2

Environment information

Environment

OS: (e.g., "Ubuntu 20.04") kernel version 5.10.233-224.894.amzn2.x86_6
Compiler(if manually compiled): (e.g., "go 14.2")

eBPF Collector configuration

#reducer configuration

#telemetry_port: 8000

num_ingest_shards: 1

num_matching_shards: 1

num_aggregation_shards: 1

partitions_per_shard: 1

enable_id_id: false

enable_az_id: false

enable_flow_logs: false

enable_otlp_grpc_metrics: false

otlp_grpc_metrics_address: "localhost"

otlp_grpc_metrics_port: 4317

otlp_grpc_batch_size: 1000

enable_otlp_grpc_metric_descriptions: false

disable_prometheus_metrics: false

shard_prometheus_metrics: false

prom_bind: "0.0.0.0:7010"

#scrape_size_limit_bytes: 0

#internal_prom_bind: "0.0.0.0:7001"

#stats_scrape_size_limit_bytes: 0

scrape_metrics_tsdb_format: "prometheus"

disable_node_ip_field: false

enable_autonomous_system_ip: true

#geoip_path: ""

enable_aws_enrichment: true

enable_percentile_latencies: false

disable_metrics: "http.all,udp.all,tcp.all,dns.all"

enable_metrics: "tcp.bytes,tcp.new_sockets,tcp.active,udp.bytes,dns.responses"

index_dump_interval: 0

Log output

Additional context

No response

[yonch]

When an RDS instance is accessed via IP, the cloud collector should supply the enrichment information. Was the cloud collector running, and did it connect successfully with the reducer?

[letian0805]

The cloud collector is running. It should not be the unknown that appears when accessing RDS, but the unknown that appears when accessing EC2 through IP. In short, if the name cannot be obtained based on the IP, should the IP be used directly?

[yonch]

the workload name is meant to be low cardinality, so only has a service name, not IP addresses which could change frequently. So (if I understand your setting and) there is no name enrichment, (unknown) would be expected.

When IP-granulatity metrics are enables, there would be an IP dimension with the IP address.