elasticluster resize - with new image_id

[Maiken Pedersen]

Hi,

I am expanding my cluster, and need to use a new image_id.

However, how do I get elasticluster to read the new values in my configuration?

I have added the new image_id in .elasticluster/config

I deactivate the virtualenv and activate it again (not sure if that is necessary).

But elasticluster remembers the old image_id and wants to use that:

2019-05-20 20:49:03 elasticluster-final.novalocal gc3.elasticluster[14318] ERROR Could not start node `compute010`: Image 4756b700-9489-4d59-bfd6-24d3b8b4167b is not active. (HTTP 400) (Request-ID: req-9026746f-324c-41a2-848f-849142146e73) -- <class 'novaclient.exceptions.BadRequest'>

I hacked my way through it changing the .elasticluster/storage/$clustername.yml file with the correct image_id for the nodes,  but I am sure there is a better way!

Maiken

[Riccardo Murri]

Hello Maiken,

> I hacked my way through it changing the .elasticluster/storage/$clustername.yml file with the correct image_id for the nodes, but I am sure there is a better way!

Unfortunately no :-(

The way to do it without hacking the `.elasticluster/storage/*.yml`
files would be to:

* add a second type of compute nodes (call it `compute2`) into the
`[setup/*]` section of the cluster, which is identical to the existing
`compute` nodes:

[setup/slurm]
frontend_groups=slurm_master
compute_groups=slurm_worker
compute2_groups=slurm_worker

* add a [cluster/*/compute2] section with the new image ID:

[cluster/slurm]
# ...everything stays the same...

[cluster/slurm/compute2]
image_id=new-image-ID

* now add nodes of class `compute2` to the cluster:

elasticluster resize -a compute2:1 slurm

I should eventually add a `elasticluster freshen` command that merges
changes in the configuration file into an existing cluster state file,
but it's a long way from where we stand now.

Ciao,
R

[Maiken Pedersen]

Hi,

thanks a lot for the answer.

But this at least is a better solution that what I was doing. So that would work. Thanks!

Maiken