Files

137 lines
4.2 KiB
YAML

extensions:
health_check:
endpoint: 0.0.0.0:13133
file_storage:
directory: /var/lib/otelcol/queue
timeout: 10s
receivers:
otlp:
protocols:
grpc: {endpoint: 0.0.0.0:4317}
http: {endpoint: 0.0.0.0:4318}
prometheus:
config:
scrape_configs:
- job_name: otel-collector
static_configs:
- targets: ["127.0.0.1:8888"]
- job_name: sms-service
scrape_interval: 15s
scrape_timeout: 10s
metrics_path: /metrics
static_configs:
- targets: ["sms-service:8080"]
labels: {service_name: sms-service}
- job_name: sms-worker
scrape_interval: 15s
scrape_timeout: 10s
metrics_path: /metrics
static_configs:
- targets: ["sms-worker:9464"]
labels: {service_name: sms-worker}
- job_name: keycloak
scrape_interval: 30s
scrape_timeout: 15s
metrics_path: /auth/metrics
static_configs:
- targets: ["keycloak:9000"]
labels: {service_name: keycloak}
metric_relabel_configs:
- action: keep
source_labels: [__name__]
regex: "(keycloak_.*|jvm_.*|process_.*|system_.*|agroal_.*)"
- action: labeldrop
regex: "(client_id|uri|session_id|user_id|phone|email)"
processors:
memory_limiter:
check_interval: 1s
limit_mib: 384
spike_limit_mib: 96
resource/common:
attributes:
- {key: service.namespace, value: han-chat, action: upsert}
- {key: deployment.environment, value: "${env:APP_ENV}", action: upsert}
- {key: service.version, value: "${env:RELEASE_VERSION}", action: upsert}
- {key: host.name, value: "${env:HOST_NAME}", action: upsert}
attributes/redact:
actions:
- {key: http.request.header.authorization, action: delete}
- {key: http.request.header.cookie, action: delete}
- {key: http.response.header.set-cookie, action: delete}
- {key: url.query, action: delete}
- {key: url.full, action: delete}
- {key: http.target, action: delete}
- {key: http.request.body, action: delete}
- {key: db.statement, action: delete}
- {key: db.query.text, action: delete}
- {key: enduser.id, action: delete}
- {key: user.phone, action: delete}
- {key: user.email, action: delete}
- {key: messaging.message.body, action: delete}
- {key: aws.s3.key, action: delete}
- {key: s3.object.key, action: delete}
filter/noise:
error_mode: ignore
traces:
span:
- 'attributes["http.route"] == "/health/live"'
- 'attributes["http.route"] == "/nginx-health/live"'
tail_sampling:
decision_wait: 10s
num_traces: 10000
expected_new_traces_per_sec: 50
policies:
- name: errors
type: status_code
status_code: {status_codes: [ERROR]}
- name: slow
type: latency
latency: {threshold_ms: 2000}
- name: successful-sample
type: probabilistic
probabilistic: {sampling_percentage: 10}
batch:
timeout: 5s
send_batch_size: 1024
send_batch_max_size: 2048
exporters:
otlp/remote:
endpoint: "${env:OTEL_REMOTE_ENDPOINT}"
headers:
authorization: ${file:/run/secrets/otel_remote_auth_header}
tls:
insecure: "${env:OTEL_REMOTE_TLS_INSECURE}"
sending_queue:
enabled: true
num_consumers: 4
queue_size: 10000
storage: file_storage
retry_on_failure:
enabled: true
initial_interval: 5s
max_interval: 30s
max_elapsed_time: 0s
service:
extensions: [health_check, file_storage]
pipelines:
traces:
receivers: [otlp]
processors: [memory_limiter, resource/common, attributes/redact, filter/noise, tail_sampling, batch]
exporters: [otlp/remote]
metrics:
receivers: [otlp, prometheus]
processors: [memory_limiter, resource/common, attributes/redact, batch]
exporters: [otlp/remote]
logs:
receivers: [otlp]
processors: [memory_limiter, resource/common, attributes/redact, batch]
exporters: [otlp/remote]
telemetry:
metrics:
address: 0.0.0.0:8888
level: normal