[ https://issues.apache.org/jira/browse/MESOS-3354?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14726064#comment-14726064 ]
Benjamin Mahler commented on MESOS-3354: ---------------------------------------- As a first step, consider exporting additional global and per-framework metrics. Granted it's not as complete as streaming out cluster events, but it's a much easier change and will give you some visibility / history (in your monitoring system). > Record scheduling statistics and history > ---------------------------------------- > > Key: MESOS-3354 > URL: https://issues.apache.org/jira/browse/MESOS-3354 > Project: Mesos > Issue Type: Improvement > Components: allocation > Reporter: Timothy Chen > > Currently there is not much visibility into why/what allocation decisions was > made, and by providing some statistics and history such as: > - Framework shares over period of time > - Allocation history > - Accept/decline history > We can have a better way to detect, and improve the scheduler, and also opens > a better way to test new implementations -- This message was sent by Atlassian JIRA (v6.3.4#6332)