Join Log in
Score:0
Hexomédia avatar Server

Incorrect ressource allocation on OpenStack

bd flag
Hexomédia

In our cluster, we have 4 computes running and we have an issue with the number 4.

We can't create VMs on it, we can't migrate VMs to or from that node. VMs are still perfectly working though. After a first diagnosis, it appears that there's a problem with the ressource provider.

Node is declared in the db with:

  • name: os-compute-4, uuid: d12ea77b-d678-40ce-a813-d8094cabbbd8

Here are the ressource provider:

  • name: os-compute-4, uuid: a9dc2a56-5b2d-49b1-ac47-6d996d2d029a
  • name: os-compute-4.openstack.local, uuid: d12ea776-d678-40ce-a813-d8094cabbbd8

In our opinion, os-compute-4.openstack.local shouldn't be there at all. We want to destroy both of the ressource provider and recreate one.

I must also precise that os-compute-4 ressource provider has 0 allocation and os-compute-4.openstack.local only 3. Moreover, their uuid doesn't correspond to any existing VMs. Overall, none of the VMs has a ressource allocation on os-compute-4.

We found the command nova-manage placement heal_allocations on the Internet but we can't find it in any container, maybe deprecated ? The cluster is running Ussuri installed with Openstack-ansible.

If you have any suggestion, any help would be appreciated. Thanks. :)

56
0 + 0
Score:0
avatar Server
us flag
Otto

Usually, it's possible to clear the resources provider from the DB and restart nova-compute service on the compute node, after a while nova will resubmit a new record of the current resource usage which would be the correct ones.

Wrt nova-mange, it should be installed on the controller node, the node running nova-api services, you can loop via ansible on all nodes and try to locate the process or check the assigned ansible roles.

You can also use the package osc-placement [1] which extend openstack cli with resources provider related subcommands.

Once installed you can issue resource provider delete command on the stale record and then restart nova-compute service on the compute node which would trigger records synchronization.

[1] https://docs.openstack.org/osc-placement/latest/index.html

0 + 0
ธงชาติไทย
Elon Musk
I sit in a Tesla and translated this thread with Ai:

mangohost

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.