Compute node send wrong nova instances number

asked 2018-03-21 20:15:20 -0600

Carlos Martins gravatar image

Recently I openned a topic about "Openstack only build newer instances on 1 compute node", after researches, I have now:

In Openstack instance building process, Nova Scheduler select the best choice of computer node to start the new instance based in weight. But I have 2 compute nodes that send number os instances as "0" as log below:

Weighed [WeighedHost [host: (node05, node05) ram: 23864MB disk: 2883584MB io_ops: 0 instances: 26, weight: 1.35044793632], WeighedHost [host: (node04, node04) ram: 11576MB disk: 2953216MB io_ops: 0 instances: 24, weight: 1.18143347491], WeighedHost

[host: (node02, node02) ram: 63803MB disk: 186368MB io_ops: 0 instances: 0, weight: 1.06310679612], WeighedHost [host: (node01, node01) ram: 53563MB disk: 69632MB io_ops: 0 instances: 0, weight: 0.863084342727], WeighedHost [host: (node03, node03) ram: 43323MB disk: 138240MB io_ops: 0 instances: 0, weight: 0.725821944816]]

We can see that node05 and node04 have correct instances number sent to controller, but nodes 03, 02 and 01 not. And both nodes 03, 02 and 01 have instances running.

The command "nova hypervisor list" show the correct instance number too. The dashboard show correct number too.

But at newer instances creation, the nova-scheduler from controller node receives wrong instances number of nodes 03, 02 and 01.

Can someone help me?

edit retag flag offensive close merge delete