This message was deleted.

We don’t dump any logs locally. The configuration you provided are to disable logging from the console when running

serve

Ok, @Aaron Pham thanks! Does logging from the console when running

serve

take some memory? I defined the following

bentoml_configuration.yml

file:

api_server:
  workers: 2
  metrics:
    enabled: false
    namespace: bentoml_api_server
  logging:
    access:
      enabled: false
      request_content_length: true
      request_content_type: true
      response_content_length: true
      response_content_type: true
      format:
        trace_id: 032x
        span_id: 016x

runners:
  onnx_mrr_runner:
    batching:
      enabled: true
      max_batch_size: 5000
      max_latency_ms: 10000
    logging:
      access:
        enabled: false
        request_content_length: true
        request_content_type: true
        response_content_length: true
        response_content_type: true
    metrics:
      enabled: false
      namespace: bentoml_runner
  onnx_mfr_runner:
    batching:
      enabled: true
      max_batch_size: 5000
      max_latency_ms: 10000
    logging:
      access:
        enabled: false
        request_content_length: true
        request_content_type: true
        response_content_length: true
        response_content_type: true
    metrics:
      enabled: false
      namespace: bentoml_runner
  onnx_mfsa_runner:
    batching:
      enabled: true
      max_batch_size: 5000
      max_latency_ms: 10000
    logging:
      access:
        enabled: false
        request_content_length: true
        request_content_type: true
        response_content_length: true
        response_content_type: true
    metrics:
      enabled: false
      namespace: bentoml_runner
  onnx_mrrr_runner:
    batching:
      enabled: true
      max_batch_size: 1000
      max_latency_ms: 10000
    logging:
      access:
        enabled: false
        request_content_length: true
        request_content_type: true
        response_content_length: true
        response_content_type: true
    metrics:
      enabled: false
      namespace: bentoml_runner
  onnx_err_runner:
    batching:
      enabled: true
      max_batch_size: 5000
      max_latency_ms: 10000
    logging:
      access:
        enabled: false
        request_content_length: true
        request_content_type: true
        response_content_length: true
        response_content_type: true
    metrics:
      enabled: false
      namespace: bentoml_runner
  onnx_errr_runner:
    batching:
      enabled: true
      max_batch_size: 25000
      max_latency_ms: 10000
    logging:
      access:
        enabled: false
        request_content_length: true
        request_content_type: true
        response_content_length: true
        response_content_type: true
    metrics:
      enabled: false
      namespace: bentoml_runner
  ct2_fp16_fr_en_runner:
    batching:
      enabled: true
      max_batch_size: 1000
      max_latency_ms: 10000
    logging:
      access:
        enabled: false
        request_content_length: true
        request_content_type: true
        response_content_length: true
        response_content_type: true
    metrics:
      enabled: false
      namespace: bentoml_runner

1. So, I can enable all

logging:access:enabled

without taking disk memory on cloud instance? 2. How could I define that I just want logging above the

warning

level on the console? 3. Does

metrics:enabled:true

takes memory space?

1. Logging should not consume storage unless you've configured the logs to be written to disk. 2. I believe there's not a way to do this at the moment with the CLI; you'll have to configure logging using the default python APIs in your service code. 3. Yes, enabling metrics will consume more memory and disk space, but it shouldn't be a very large overhead.

Thanks @sauyon! Before release 1.0 of BentoML there was an environment variable to set logging to warning level and above with:

env:
          - name: BENTOML__LOGGING__LEVEL
            value: "WARN"

Is this environment variable still valuable with the new release?

No, but you should be able to set the config options with environment variables, see https://docs.bentoml.org/en/latest/guides/configuration.html#providing-configuration-during-serve-runtime.