SSH/SCP/SFTP protocol support #267
Conversation
| # | ||
| # Global storage for SSH connections. | ||
| # | ||
| _SSH = {} |
There was a problem hiding this comment.
How does such global cache play with multiprocessing / forking?
There was a problem hiding this comment.
For multiprocessing: no problem, there will be one cache for subprocess.
For forking: I think it will be a problem. We don't want different threads to access the same connection simultaneously. Perhaps we can include the current thread_id in the cache key?
There was a problem hiding this comment.
I don't think we have any solid contract in place for concurrency (yet).
My naive expectation as a user would be "A 'file' open in one process/thread is always valid for that process/thread, but not necessarily across process or thread boundaries". But we should probably document that explicitly.
How do the other connection types (S3, HTTP, HDFS…) manage in the face of multiple processes or threads?
There was a problem hiding this comment.
OTOH they "should be fine". We haven't had any concurrency-related bug reports.
On the other hand, we don't have any explicit testing for concurrency problems, either.
Co-Authored-By: mpenkov <penkov@pm.me>
Continued from #58 cause I couldn't write to the original author's repo.
Fix #36