Open ubergarm opened 5 years ago
It's marked as a known issue in the release notes for 0.2.x. The problem is not fixing the return code bug it presently suffers from, or finding a neater way to grant authority between the two nodes, but that the implementation is missing a super important feature -- flow control.
Without this, unbounded RAM growth can occur in a pipelined sender like rsync.
Flow control is a solved problem elsewhere -- services.FileService knows how to do it. I want to extract that mechanism, blend it with the 'out of band control' bits of fakessh, and produce a new abstraction that can be used in both places and generically -- named something like Pipe.
I'm getting close to looking at this stuff again -- it's needed for fully generic tunnelling Ansible synchronize
module, but the practicalities of keeping bug count down for extension users is a little too high :) The great news is that this Ansible work has flushed out a megaton of bugs in the core library, it's just taking a little more time than anticipated to reach the end goal
At the very least I need to add a caution box on the index page in the SSH section. Sorry about that!
Great, thanks for clarifying!
Hi @dw !
I'm trying to use Mitogen not as part of Ansible, but as a standalone solution to connect to remotes, copy files in/out, and execute Python code.
As part of this work, I tried to use the fakessh functionality of Mitogen to reuse an existing SSH context to run rsync
commands, but it fails.
I get that fakessh is pretty much unmaintained at this moment. Just wanted to ask, are there any plans to fix that soon? I'm willing to help fixing its shortcomings.
Is
fakessh
broken? It seems skipped in the unittestMy goal is to run
git
on a remote server that uses my local laptop'sssh
context. I do not want to forward my ssh auth with-A
.Is this possible currently or in the future? Thanks for clearing this up, couldn't find any full working examples...
My failed attempt w/
Python 3.6.6
andmitogen 0.2.3