Not sure this is supported at all, but I'm quite sure this is not a good idea: in such a configuration if you run a DeployJob for a 10 MB payload against 100 targets, it means the staging directory will need to hold at least 1GB. Performance wise it's also probably a bit challenging: 100 simultaneous writes to the same folder will certainly reach a bottleneck at some point.
also - if you are deploying the same payload to multiple targets at once, then each target would need a unique directory on the nfs location or else the staging would clobber each other.
what is the issue using a local directory ?
Yes that's true too. That's why we introduced the ability to deploy to Components, which obviously would be a bit cumbersome if a strict requirement for all deployments..
i'm not sure how that's relevant - even deploying to components will use the STAGING_DIR value which is not unique. i don't understand why a nfs location would be preferable to a local location...
I don't quite get it either.
I mean using a separate filesystem for staging is fine if you want to prevent uncontrolled deployments to harm the O/S or the agent, but as you said, local.
NFS only makes sense in the context of the agent if you want to have several servers accessing the same data at the same time (i.e. multiple appservers accessing the fileserver path) or don't want to locally copy sources during a deployment (i.e. patches or large installable files like Oracle)