Project

General

Profile

Task #2705

Story #2036: Stress testing of components

Ensure replication queue can handle large numbers of pending tasks

Added by Matthew Jones about 9 years ago. Updated almost 9 years ago.

Status:
Closed
Priority:
Normal
Assignee:
Category:
d1_replication
Start date:
Due date:
% Done:

100%

Milestone:
CCI-1.0.0
Product Version:
*
Story Points:
Sprint:

History

#1 Updated by Matthew Jones about 9 years ago

  • Category set to d1_replication

When MNs come online, a large bolus of objects may need to be replicated -- up to approximately a few hundred thousand objects for the KNB, for example. Need to ensure that the rep task queues in hazelcast can handle this number of objects times the number of replicas, so probably 3x objects in the rep queue until they work their way out.

#2 Updated by Chris Jones almost 9 years ago

  • Status changed from New to Closed

We're using the hzReplicationEvents queue and the hzReplicationTasks queue to buffer backlogged events and tasks. These structures hold small objects, so memory shouldn't be an issue for 300K entries.

Also available in: Atom PDF

Add picture from clipboard (Maximum size: 14.8 MB)