Cloudera Manager 3.7.4 released! by Bala Venkatrao.
From the post:
We are pleased to announce that Cloudera Manager 3.7.4 is now available! The most notable updates in this release are:
- A fixed memory leak in supervisord
- Compatibility with a scheduled refresh of CDH3u3
- Significant improvements to the alerting functionality, and the rate of ‘false positive alerts’
- Support for several new multi-homing features
- Updates to the default heap sizes for the management daemons (these have been increased).
The detailed Cloudera Manager 3.7.4 release notes are available at: https://ccp.cloudera.com/display/ENT/Cloudera+Manager+3.7.x+Release+Notes
Cloudera Manager 3.7.4 is available to download from: https://ccp.cloudera.com/display/SUPPORT/Downloads
I admit to being curious (or is that suspicious?) and so when I read ‘false positive alerts’, I had to consult the release notes:
- Some of the alerting behaviors have changed, including selected default settings. This has streamlined some of the alerting behavior and avoids spurious alerts in certain situations. These changes include:
- The default alert values have been changed so that summary level alerts are disabled by default, to avoid unnecessary email alerts every time an individual health check alert email is sent.
- The default behavior for DataNodes and TaskTrackers is now to never emit alerts.
- The “Job Failure Ratio Thresholds” parameter has been disabled by default. The utility of this test very much depends on how the cluster is used. This parameter and the “Job Failure Ratio Minimum Failing Jobs” parameters can be used to alert when jobs fail.
So, the alerts in question were not spurious alerts but alerts users of Cloudera Manager could not correctly configure?
Question: Can your Cloudera Manager users correctly configure alerts? (That could be a good Cloudera installation interview question. Use a machine disconnected from your network and the Internet for testing.)