riseml/loghandler repository overview

⁠RiseML Log Handler

The RiseML log handler reads experiment logs and states from a queue and stores the logs on disk. It provides a websocket endpoint to stream logs and state updates to clients.

⁠List of used environment variables:

PORT
Express.js webserver port.
If no value value is provided, 3200 is used.

LOG_DIR
Directory where to store job logs. Example: /log

RISEML_API_URL
RiseML API URL.
Example: http://api:3000

RISEML_APIKEY
Token to authenticate loghandler's requests to RiseML API.
Example: admin_key

AMQP_URL
RabbitMQ amqp:// url, no default value.
Example: amqp://192.168.99.100:30567

STATE_EXCHANGE
AMQP exchange name, where state updates are broadcasted.
Example: riseml-states

LOG_QUEUE
AMQP queue name for consuming job logs, produced by k8s. Example: consolelog

MONITOR_EXCHANGE_PREFIX
Utilization AMQP exchange name prefix. Resulting exchange name will be (MONITOR_EXCHANGE_PREFIX + jobId).
Example: monitor-

NODE_ENV
Environment to run loghandler in. Can be 'development', 'test', or 'production'. The latter one activates special behavior such as sending stack traces to Rollbar. Default is 'development'. Example: development

CLUSTER_ID
The current cluster id. Used to send errors to error tracker. Example: 185015de-ae50-43e5-adc9-a7f0a36ce15b

RISEML_ROLLBAR_ENDPOINT
URL where stack traces should be send to by Rollbar's error tracking package. If not set, the fallback (https://api.riseml.com/errors/loghandler) is used. The only reason to change this is changing/developing the error tracking logic. Example: http://backend:3001/errors/loghandler