You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "Joseph K. Bradley (JIRA)" <ji...@apache.org> on 2018/04/03 18:00:00 UTC
[jira] [Created] (SPARK-23859) Initial PR for Instrumentation
improvements: UUID and logging levels
Joseph K. Bradley created SPARK-23859:
-----------------------------------------
Summary: Initial PR for Instrumentation improvements: UUID and logging levels
Key: SPARK-23859
URL: https://issues.apache.org/jira/browse/SPARK-23859
Project: Spark
Issue Type: Sub-task
Components: ML
Affects Versions: 2.3.0
Reporter: Joseph K. Bradley
Assignee: Weichen Xu
This is a subtask for an initial PR to improve MLlib's Instrumentation class for logging. It will address a couple of issues and use the changes in LogisticRegression as an example class.
Issues:
* The UUID is currently generated from an atomic integer. This is a problem since the integer is reset whenever a persisted Estimator is loaded on a new cluster. We should just use a random UUID to get a new UUID each time with high probability.
* We use both Instrumentation and Logging to log stuff. Let's standardize around Instrumentation in MLlib since it can associate logs with the Estimator or Transformer which produced the logs (via a prefix with the algorithm's name or UUID).
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org