Interacting with KBase devops people, I'm reminded of the following.
When a Globus transfer encounters an error of the sort that causes a DTS transfer to "fail" (missing file, perms problem, ???), Globus may still continue that transfer to "completion" (minus error-related casualties). This isn't what we want. DTS should issue a request to cancel the transfer upon its definition of "failure," and should try to clean up the disk so that a failed transfer doesn't leave a bunch of rubble around.
Interacting with KBase devops people, I'm reminded of the following.
When a Globus transfer encounters an error of the sort that causes a DTS transfer to "fail" (missing file, perms problem, ???), Globus may still continue that transfer to "completion" (minus error-related casualties). This isn't what we want. DTS should issue a request to cancel the transfer upon its definition of "failure," and should try to clean up the disk so that a failed transfer doesn't leave a bunch of rubble around.