Details
-
Bug
-
Status: Closed (View Workflow)
-
Minor
-
Resolution: Fixed
-
None
Description
In the code that calculates how many nodes we should scale up to based on queue depth, it assumes that all slaves just have one executor but this is user configurable so if you give your slaves a moderate number of executors, this plugin can over-scale by a large factor.
Attachments
Activity
Erik Lattimore
created issue -
Erik Lattimore
made changes -
| Field | Original Value | New Value |
|---|---|---|
| Summary | This plugin scales up too aggressively when your have nodes with more than one executor | This plugin scales up too aggressively when you have nodes with more than one executor |
Erik Lattimore
made changes -
| Status | Open [ 1 ] | In Progress [ 3 ] |
Mark Lagendijk
added a comment - Can this pull request be merged in please?
Because of this issue we have to set 'max' to a low number. Whenever scaling up is needed it will often unnecessarily scale to the max number of nodes.
Mark Lagendijk
added a comment - Can this pull request be merged in please?
Because of this issue we have to set 'max' to a low number. Whenever scaling up is needed it will often unnecessarily scale to the max number of nodes. Erik Lattimore
added a comment - This is fixed in 1.1.4 Erik Lattimore
made changes -
| Resolution | Fixed [ 1 ] | |
| Status | In Progress [ 3 ] | Closed [ 6 ] |
See https://github.com/jenkinsci/ec2-fleet-plugin/pull/5 for my attempt at fixing the problem.