[Known issue] "There are not enough hosts available" or "No valid host was found" : ARDC Support

Start a new topic

[Known issue] "There are not enough hosts available" or "No valid host was found"

Stephen Crawley

started a topic over 3 years ago

When you attempt to launch a new instance in the Nectar cloud, the launch may fail with the error messages "There are not enough hosts available" or "No valid host was found" This is a fairly common occurrence.

What does it mean?

The normal meaning of this message is that the availability zone that you selected is "full". The selected AZ does not have enough free resources to launch an instance of the size that you requested. If you didn't specifically select an AZ, it could mean none of the AZs that you have access to have enough free resources.

It can also mean that you have selected an AZ where the resources are restricted to particular classes of project, and your project is not "tagged" appropriately. For example, University of Melbourne currently have a UoM private AZ that is restricted to UoM local projects. If you attempt to launch there and your project is not tagged, you will get the "not enough hosts" error.

It typically does not mean that Nectar, or a specific Nectar node is broken, though node problems or operational restrictions may cause this on occasion.

What can I do about this?

The general approach is to try again:

Try again. This could be a transient issue.
Try again later. Someone may have deleted some instances, freeing up capacity for new ones to be created.
Try again with a smaller flavor. Try a flavor with fewer VCPUs or less RAM. Try an 'm3' flavor rather than 'm1' or 'm2. (Some AZs have no free capacity for instances of older flavors.)
Try again in a different AZ. This may not be an option for users with a "local funded" allocation. Also, it may not be an option if your instance needs to use storage (e.g. a volume) that is tied to a particular Nectar node.
If you didn't select a specific AZ, try selecting one.

If trying again doesn't help, raise a support request. They will typically pass your request to the Node operators to see if they can free up some capacity. However, you may be disappointed, or you may have to make do with a smaller instance than you wanted.