Like most systems, ScienceOps has logs. The logs are split up into 2 categories:

  • Model Logs: Logs corresponding to any output produced by the user's models.
  • System Logs: Logs produced by the ScienceOps system.

Model Logs

These logs can be accessed via either the web application or by inspecting the /var/scienceops/modellogs directory on the master node.

The web application logs are handy for data scientist or other users deploying models. They are not intended to be production-grade logs for investigating system level issues. Web application logs are also limited to the last 1000 lines of the log file.

Model logs stored on disk are accessible via SSH (or likewise) access in the /var/scienceops/modellogs directory. Log files are stored as line-delimited JSON (.jsonl) files:

{"InstanceId":3159,"User":"kermit","Model":"HelloWorld","Timestamp":"2016-09-22T15:13:53.40797417Z","hello world","DeploymentId":365}
{"InstanceId":3159,"User":"kermit","Model":"HelloWorld","Timestamp":"2016-09-22T15:13:54.326402089Z","this is some log data","DeploymentId":365}
{"InstanceId":3159,"User":"kermit","Model":"HelloWorld","Timestamp":"2016-09-22T15:13:55.316944806Z","this is also some log data","DeploymentId":365}

System Logs

Logs for the actual ScienceOps process can be found on the master and worker nodes. The location of the logs depends on the opeating system you're running:

  • RHEL 6 / CentOS6: /var/log/scienceops/
  • RHEL7 / CentOS7: journalctl -u scienceops (master), journalctl -u scienceops-worker (worker)
  • Ubuntu 14.04: /var/log/scienceops/

results matching ""

    No results matching ""