Instance failover on Nova-compute node failure

asked 2015-10-05 04:59:29 -0600

nd gravatar image

The context of the question is - how do we recover an instance with boot disk on persistent storage from Nova node failure.

The obvious requirement would be to have shared storage which is visible from multiple Nova nodes.

Now suppose we use iSCSI / FC based storage connectivity [e.g. using EMX VNX series / NetApp Storage array], from Cinder and Nova nodes. Assume that the instance is created/booted with a persistent volume that cinder created in the backend Storage array.

My question is - when a volume is created by tenant/cinder - is the LUN created in the backend Storage array mapped to multiple Nova nodes?

If not how can the instance be started on a different Nova node incase the Nova node running the instance fails?

Thanks, nd

edit retag flag offensive close merge delete