Use of OpenStack Availability zone

[Alex Loew]

How can I specify in the config file, that Elasticluster should use a specific availability zone for the rollout of the cluster?

[Riccardo Murri]

Hello Alex,

> How can I specify in the config file, that Elasticluster should use a specific availability zone for the rollout of the cluster?

Which cloud provider/backend?

Kind regards,
Riccardo

[Alex Loew]

sorry, Openstack

[Riccardo Murri]

Hi Alex,

ElastiCluster currently has no code to place nodes in a specific
availability zone on OpenStack clouds.
It would not be difficult to add it, if there is request.

I'm curious, however, as to what the use case for availability zones
would be, when using ElastiCluster. In my understanding, AZs are used
to spread out resources for high-availability: e.g., you would put the
main DB and its replica in different AZs so that if one zone fails,
the other one has good chances to be still up. But in the case of a
compute cluster?

Ciao,
R

[Alex Loew]

Riccardo,

many thanks for the answer. We are running a multi site openstack cloud with 3 AZ. I have got 2 hypervisors for the HPC cluster and they are in a specific AZ. In order to setup the Elasticluster I would need to specify which AZ to use. This is the problem I have at the moment, as I cannot influence the scheduler to put my requests to the correct AZ. The machines get startet and end in error, because the resources bound to the flavour cannot be found in this AZ.

How can I raise this feature request?

Many thanks for the support!

AL

[Riccardo Murri]

Hello Alex,

many thanks for the detailed explanation! Can you please submit a
bug/feature request at
https://github.com/gc3-uzh-ch/elasticluster/issues/new ?

FWIW, we use dedicated HPC flavors (as opposed to AZs) to accomplish
the same goals on our internal OpenStack cloud.

Ciao,
R