elasticluster resize - remove all compute nodes

[Maiken Pedersen]

Hi again,

so I discovered that it does currently not work to remove all compute nodes with the resize command in elasticluster.

Since there are no more compute nodes (or slurm_workers) left when I have issued the resize command removing all the slurm workers, I get this:

ASK [nfs-server : Ensure export directories exist] ******************************************************************************************************************
fatal: [frontend001]: FAILED! => {"msg": "'dict object' has no attribute 'slurm_worker'"}

It might very well be that other tasks would fail for the same reason, but elasticluster quits after this.

Maiken

[Riccardo Murri]

Hello Maiken,

> so I discovered that it does currently not work to remove all compute nodes with the resize command in elasticluster.

Related: https://github.com/gc3-uzh-ch/elasticluster/issues/248

> Since there are no more compute nodes (or slurm_workers) left when I have issued the resize command removing all the slurm workers, I get this:
>
> ASK [nfs-server : Ensure export directories exist] ******************************************************************************************************************
> fatal: [frontend001]: FAILED! => {"msg": "'dict object' has no attribute 'slurm_worker'"}

In principle this is can be fixed by checking that a group exists
before listing its members but... what is the purpose of having a
SLURM cluster with no worker nodes?

Ciao,
R

[Maiken Pedersen]

:)

So my situation is that all the compute nodes in my cluster must be removed, and new ones must be added.

But I think I will just destroy the cluster and start with a new one. And just replace the frontend with the old frontend machine ones the new cluster is set up.

Seems a bit easier to accomplish.

Maiken