Bug #12199

Updated by Peter Amstutz about 2 years ago


* API server gains a node_sizes table with fields:
** uuid
** cloud instance id
** price
** cores
** ram
** disk
* When creating a container record from a container request, API server determines determine the appropriate node size and puts put the uuid in "node_size" field under "scheduling_parameters"
* In crunch-dispatch-slurm, look at "node_size" under "scheduling_parameters" and add it to the sbatch command line in "--constraints"
* In nodemanager, for each slurm job with constraints, consult the API server's "node_types" table to get the corresponding cloud node type for the wishlist.
* The "Size" sections of the node manager configuration may be ignored in favor of the information from the API server's node_types.

Back