Task #2624
Story #2622: d1_replication should prioritize MN replication tasks based on load, failures, and bandwidth factors
Modify ReplicationManager.createAndQueueTasks() to limit replication tasks based on current MN load
100%
Description
As a first pass on trying to throttle replicate() calls to possibly bogged-down MNs, change createAndQueueTasks() such that we evaluate how many currently pending tasks there are for a given node, and cap the calls based on a configuarable limit.
See the epad for algorithm details:
History
#1 Updated by Chris Jones almost 12 years ago
- Status changed from New to In Progress
Added the prioritizeNodes() method to implement the new prioritization algorithms. Needs testing.
#2 Updated by Chris Jones almost 12 years ago
- Status changed from In Progress to Closed
Added the getRequestFactors(), getFailureFactors(), and getBandwidthFactors() methods that feed into prioritizeNodes(). Skye moved these methods into a new ReplicationPrioritizationStrategy class for modularity and independent testing. Testing showed that the prioritization scheme works fine now that we give a 5 try grace period for MNs before calculating scores.
