Some Protocols execute tools that emit log information to either STDOUT, STDERR, or both. The information in these streams can be critical for users to understand where their Tasks are going wrong in case of persistent failure.
We should capture these streams when compute services execute ProtocolDAGs, then preserve them as additional artifacts when submitting completed or failed ProtocolDAGResults back to the server. These log artifacts can be stored in the object store, with references in the state store.
We can then expose these artifacts via additional AlchemiscaleClient methods, such as:
get_task_stdout(task_sk) : return all stdout outputs from all ProtocolDAGResults associated with the given Task
get_task_stderr(task_sk) : return all stderr outputs from all ProtocolDAGResults associated with the given Task
Some
Protocol
s execute tools that emit log information to either STDOUT, STDERR, or both. The information in these streams can be critical for users to understand where theirTask
s are going wrong in case of persistent failure.We should capture these streams when compute services execute
ProtocolDAG
s, then preserve them as additional artifacts when submitting completed or failedProtocolDAGResult
s back to the server. These log artifacts can be stored in the object store, with references in the state store.We can then expose these artifacts via additional
AlchemiscaleClient
methods, such as:get_task_stdout(task_sk)
: return all stdout outputs from allProtocolDAGResult
s associated with the givenTask
get_task_stderr(task_sk)
: return all stderr outputs from allProtocolDAGResult
s associated with the givenTask