Instance can't be resized to a different flavor when all hypervisor CPUs are used.
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Mirantis OpenStack |
Invalid
|
Medium
|
Olga Klochkova |
Bug Description
Hypervisor has enough memory to perform resize
When active instance is being resized to a different flavor having the same amount of CPUs but different amount of RAM, operation fails with error message: "Error: No valid host was found. No valid host found for resize (HTTP 400) (Request-ID: req-0837fc41-
Expected result: since amount of used CPUs should not increase as a result of the resize, operation should succeed.
VERSION:
feature_groups:
- mirantis
production: "docker"
release: "8.0"
api: "1.0"
build_number: "569"
build_id: "569"
fuel-nailgun_sha: "558ca91a854cf2
python-
fuel-agent_sha: "658be72c4b42d3
fuel-
astute_sha: "b81577a5b7857c
fuel-library_sha: "33634ec27be77e
fuel-ostf_sha: "3bc76a63a9e7d1
fuel-mirror_sha: "fb45b80d7bee58
fuelmenu_sha: "78ffc73065a967
shotgun_sha: "63645dea384a37
network-
fuel-upgrade_sha: "616a7490ec7199
fuelmain_sha: "d605bcbabf3153
tags: | added: nova |
description: | updated |
Changed in mos: | |
status: | Incomplete → Invalid |
Olga, could you please provide nova-scheduler logs?
Overall, I think the behaviour you are seeing is valid in the sense, that resize effectively creates a new VM, thus, it has to check if there are enough resources for that at the time when the check is actually performed, i.e. before the original VM is deleted.
So, IMO, this should probably become Invalid.