Details
-
Improvement
-
Status: Open
-
Minor
-
Resolution: Unresolved
-
None
-
None
-
None
Description
Currently we either accept or decline every offer of each bath of resourceOffers() to acquire a fresh set of offers ASAP.
This may be a good practice in the perspective that Mesos Master can quickly acknowledge that the job does not need the resources and make better cluster-wide decisions based on that.
But on the other hand this may put too much pressure on Mesos Master. I noticed a frequent exchange of new offers & offer declination in the job duration.
It would be good to study the approaches other frameworks(especially those for long running jobs such as Aurora, Marathon, and Spark) take.