Feature #2881: [OPS] Basic node manager that can start/stop compute nodes based on demand
fix update_node_attributes script and/or merge its remaining functionality into crunch-dispatch
this script updates the node objects with information about what is running on them, which us information used by node manager
Non-existant node object is not a fatal state; node manager will create
those objects as needed.
crunch_dispatch: when a Node object state is changed to not alloc/comp,
make sure to change the running_job_uuid field to nil.
#2 Updated by Brett Smith almost 6 years ago
node.info[:running_job_uuid]. We would rather start expressing this as a relationship, each Node has one Job, so that querying Nodes will use the existing permission code to determine access to the Job UUID. Job queries can join with the Node table to return the full set of Nodes running it.
This seems to be the only functionality left that update_node_attributes currently does that crunch-dispatch doesn't.