Details
-
Improvement
-
Status: Resolved
-
Major
-
Resolution: Duplicate
-
None
-
None
-
None
Description
When running Mesos in multi-tenant environment with several roles and frameworks, it's easy to be tricked that you have enough slack capacity. This leads to unexpected delays in scheduling, since you are low on resources on specific roles and unused chunks of resources are too small.
While this is also a monitoring issue, I think Mesos UI can do a better job of vusualizing what resources are available for roles. I made a CLI tool to do that for myself: https://github.com/bobrik/scrappy
In addition, it could be worth showing top available resource chunks per resource:
- Biggest CPU intensive task that can be scheduled (with mem and disk)
- Biggest mem intensive task that can be scheduled (with cpu and disk)
- etc
These things can be monitored and alerted on, but it probably isn't the best solution. Maybe it's a job for a separate service.
Attachments
Issue Links
- duplicates
-
MESOS-6447 Display role weight / role quota information in the webui.
- Resolved