Live migration of instances requires the destination AND the source to have the same resources free on top of what the instance is taking

Bug #1975490 reported by Diko Parvanov
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
OpenStack Compute (nova)
Expired
Undecided
Unassigned

Bug Description

A live-migration of a VM with 8 vCPUs from host SOURCE to host DEST, where host SOURCE has no free vCPUs will fail, because it will require that host SOURCE has 8 vCPUs together with the already allocated 8 vCPUs from the running instance, so it requires the source host to have 16 vCPUs.

The workaround is to increase the allocation ratio of the host SOURCE via openstack resource provider inventory class set --allocation_ratio and then scale it back down to the required value.

Revision history for this message
sean mooney (sean-k-mooney) wrote :

currently we do not have enough info to traige this so setting to incomplete

what version of OpenStack was this observed with.
i believe live migration should be using multiple allocations in placement
one for the source host and a second of the dest using the migration uuid

Changed in nova:
status: New → Incomplete
tags: added: libvirt live-migration placement
Revision history for this message
Diko Parvanov (dparv) wrote :

cloud is focal/ussuri, it is using multiple allocations, but nova should not require the source host to have twice the machine of available resources

Changed in nova:
status: Incomplete → New
Revision history for this message
Balazs Gibizer (balazs-gibizer) wrote :

I tried to reproduce the described bug but I was not able to. See [1].

Please note that it is a known behavior that if the source compute is _overallocated_ (i.e. used > total*allocation_ratio - reserved) then you cannot move the VM out from that host. But If the source compute is just full, then you should be able to migrate the VM as [1] shows.

Could you please provide:
* openstack resource provider inventory list <source_compute_node_uuid>
* openstack resource provider inventory list <dest_compute_node_uuid>
* openstack resource provider usage show <source>
* openstack resource provider usage show <dest>
* openstack resource provider allocation show <vm uuid>
* openstack server event list <vm uuid>

Also please provider the relevant nova and placement logs please.

Setting this as Incomplete as I cannot reproduce the issue.

[1] https://review.opendev.org/c/openstack/nova/+/845168

Changed in nova:
status: New → Incomplete
Revision history for this message
Launchpad Janitor (janitor) wrote :

[Expired for OpenStack Compute (nova) because there has been no activity for 60 days.]

Changed in nova:
status: Incomplete → Expired
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.