[MESOS-3157] Only perform periodic resource allocations.

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Won't Fix
• Component/s: allocation
• Epic Link: allocator performance phase 2
• Assignee: Unassigned
• Reporter: James Peach
• Shepherd: Benjamin Mahler
• Created: 27/Jul/15 21:05
• Updated: 07/Jun/18 21:52
• Resolved: 07/Jun/18 21:52

Description

Our deployment environments have a lot of churn, with many short-live frameworks that often revive offers. Running the allocator takes a long time (from seconds up to minutes).

In this situation, event-triggered allocation causes the event queue in the allocator process to get very long, and the allocator effectively becomes unresponsive (eg. a revive offers message takes too long to come to the head of the queue).

We have been running a patch to remove all the event-triggered allocations and only allocate periodically on the allocation interval. This works great and really improves responsiveness.

Issue Links

• contains: MESOS-2285 Eliminate dependency on master::Flags in Allocator
• duplicates: MESOS-4766 Improve allocator performance.
• duplicates: MESOS-4767 Apply batching to allocation events to reduce allocator backlogging.
• is related to: MESOS-4102 Quota doesn't allocate resources on slave joining.
• is related to: MESOS-3353 generic mechanism to smuggle allocation options
• is related to: MESOS-4694 DRFAllocator takes very long to allocate resources with a large number of frameworks
• is superceded by: MESOS-6904 Perform batching of allocations to reduce allocator queue backlogging.
• relates to: MESOS-3078 Recovered resources are not re-allocated until the next allocation delay.