AgentError, unknown response from agent on Windows with 64-bit PuTTY about vagrant

[jaraco]

Also, please let me know if you have any recommendations for working around the issue. I'm going to downgrade to Vagrant 1.0.x in the meantime.

[jaraco]

I should mention that the issue may not be related to the word size, but could be related to UAC (communicating with the PAgeant requires special handling when the PAgeant and the requesting process are in different administrative contexts).

Indeed, I just confirmed that the error doesn't occur if I run 'vagrant up' in a non-elevated context (same as what PAgeant runs). Therefore, the issue appears to have the same cause as Paramiko Issue #98, the resolution of which required extensive use of the underlying Windows APIs to address the issue properly (as PuTTY implements it).

[mitchellh]

Could you report a bug upstream in net-ssh? https://github.com/net-ssh/net-ssh/issues?state=open

This is where the bug is and it'd be best serviced there. Thanks!

[jaraco]

I've reported the issue upstream, but the issue still remains in Vagrant. I find that I can work around the issue by stopping my agent process, but that's far from ideal. Perhaps there's something Vagrant can do to work around the issue until it's resolved in net-ssh? For example, since Vagrant already has the key, it doesn't really need to use the agent at all. Perhaps Vagrant could disable agent support, at least for the management processes so it runs like it did in 1.0?

[jaraco]

This continues to be an issue for me. Does net-ssh give the ability to disable agent support?

[zodeus]

+1, having to shutdown pageant every time is quite a nuisance.

[aljinovic]

+1, really annoying

[tschellenbach]

+1 please fix this, can't use vagrant anymore when using pageant

[tschellenbach]

the only workable solution i found so far is to use vagrant with osx or ubuntu instead of windows.
a few people in my team are on windows though and this is making vagrant quite unworkable.
are there any known workarounds?

[nathanlenz]

+1 Also experiencing this issue.

[jaraco]

What I do is (from powershell) "kill -name pageant" and I also have an alias to start-pageant with the same parameters that it starts up normally. Then I invoke those before and after running vagrant. I guess one could wrap the vagrant in an alias that does all of that automatically.

[mitchellh]

I'm sorry but if there is a way Vagrant can work around this, I'd be happy to put that in. At the moment, it doesn't sound like there is one. This seems to happen even if Vagrant explicitly tells net-ssh to NOT even care about agents.

[crisu83]

I managed to fix this my simply closing pageant, thanks for pointing me in the right direction @jaraco.

[ecki]

When I change net-ssh-2.6.8/lib/net/ssh/authenticatiob/agent/socket.rb to raise "AgentNotAvailable" in all error cases (line 79, 81) then KeyManager will happily ignore the failing/incompatible/inaccessible agent.

I sent an upstream pull request, but it might need some different startegy to make it generic: net-ssh/net-ssh#129

[delano]

The fix (from net-ssh/129) will be included in the 2.8.0 release later today.

[mitchellh]

@delano oh my god amazing. <3 <3 <3

[delano]

This is now a thing that happened:
http://rubygems.org/gems/net-ssh/versions/2.8.0

FYI, there were a lot of fixes and additions in this release so if you find any new issues let me know.

[ecki]

Yeah, my first fix and I do not even speak Ruby :)

[dopry]

@mitchellh any chance you've gotten around to testing a 1.4.4 with net-ssh bumped to <= 2.8.0? I tried locally on my windows, but many of the tests are failing. Any special tricks to running tests on windows?

[delano]

Just as an FYI, there is an issue with password auth in 2.8.0 that affects some configurations. I recommend waiting for that to be resolved before bumping the net-ssh version for vagrant.

See net-ssh/#145 for more info.

[mitchellh]

@delano Thanks, I'll wait for a 2.8.1

[Bup3]

+1

Ok i wait, too :/