juju picking randomly from available machines instead of best fit
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| | juju |
Medium
|
Unassigned | ||
| | juju-core |
Medium
|
Unassigned | ||
| | juju-deployer |
Undecided
|
Unassigned | ||
Bug Description
If I have a deployer service config like the following:
services:
B:
constraints: mem=8192M
A:
constraints: mem=1024M
And then do:
juju add-machine --constraints=
juju add-machine --constraints=
I end up with a 8GB machine 1, and 1GB machine 2. Now, running juju-deployer over the deployment config pushes service A to machine 1 (8GB) and spawns a third instance with 8GB of RAM.
I had expected juju would attempt to place each service on the smallest instance it could find that met the constraints. In this circumstance, I would expect it to use the two machines provided that exactly match the constraints.
Running juju add-machine in the order expected by deployer works around this issue.
juju-core 1.20.1
juju-deployer 0.4.3
| tags: | added: ubuntu-engineering |
| description: | updated |
| Changed in juju-core: | |
| importance: | Undecided → High |
| status: | New → Triaged |
| tags: | added: constraints deployer |
| Changed in juju-core: | |
| importance: | High → Medium |
| summary: |
- juju/deployer placement dependent on ordering + juju picking randomly from available machines instead of best fit |
| Changed in juju-core: | |
| milestone: | none → 1.21 |
| importance: | Medium → High |
| Changed in juju-core: | |
| milestone: | 1.21 → 1.22 |
| Changed in juju-core: | |
| milestone: | 1.22-alpha1 → 1.23 |
| Changed in juju-core: | |
| importance: | High → Medium |
| milestone: | 1.23 → none |
| Changed in juju: | |
| status: | New → Triaged |
| importance: | Undecided → Medium |
| milestone: | none → 2.1.0 |
| Changed in juju-core: | |
| status: | Triaged → Won't Fix |
| Changed in juju: | |
| milestone: | 2.1-rc2 → none |
primarily a core bug, marking invalid for deployer.