[MESOS-10015] updateAllocation() can stall the allocator with a huge number of reservations on an agent.

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.5.3, 1.6.2, 1.7.2, 1.8.1, 1.9.0
• Fix Version/s: 1.7.3, 1.8.2, 1.9.1, 1.10.0
• Labels: resource-management
• Target Version/s: 1.7.3, 1.8.2, 1.9.1
• Sprint: Resource Mgmt: RI-20 58
• Story Points: 5
• Assignee: Andrei Sekretenko
• Reporter: Andrei Sekretenko
• Created: 21/Oct/19 17:57
• Updated: 31/Oct/19 00:36
• Resolved: 30/Oct/19 23:59

Description

Currently, updateAllocation() called for a single-object Resources for a single framework on a single slave requires `(total number of frameworks) * (number of resource objects per this slave)^2` calls of `Resource::addable()`

In a cluster with a large number of frameworks this results in severe degradation of allocator performance when a bunch of RESERVE/UNRESERVE operations occurs for an agent with hundreds of unique resources.

On our testing cluster task we observed task scheduling delays up to 30 minutes due to allocator being occupied with processing UNRESERVE operations.

Issue Links

• is related to: MESOS-9942 Deprecate framework sorter.
• relates to: MESOS-10016 Add a benchmark for HierarchicalAllocatorProcess::updateAllocation()