• chore: change config-ssh Call to Action to use suffix #17445
• feat: modify config-ssh to check for Coder Connect #17419 👈(View in Graphite)
• feat: add coder connect exists hidden subcommand #17418
• main

This stack of pull requests is managed byGraphite. Learn more aboutstacking.

Edit: Just realized that setting-o options are only effective for proxied commands. It'd be nice to have two entries in the SSH config, i.e:

Host *.coder        Foo=BarMatch host *.coder !exec "/Users/maf/Code/coder/coder connect exists %h"        ConnectTimeout=0        StrictHostKeyChecking=no        UserKnownHostsFile=/dev/null        LogLevel ERROR        ProxyCommand /Users/maf/Code/coder/coder --global-config "/Users/maf/Library/Application Support/coderv2" ssh --stdio --hostname-suffix coder %h

Ok, now looks like

Host *.coder        ConnectTimeout=0        StrictHostKeyChecking=no        UserKnownHostsFile=/dev/null        LogLevel ERRORMatch host *.coder !exec "C:\Users\micha\repos\coder\coder.exe connect exists %h"        ProxyCommand C:\Users\micha\repos\coder\coder.exe --global-config C:\Users\micha\AppData\Roaming\coderv2 ssh --stdio --hostname-suffix coder %h

Merge activity

• Apr 17, 4:04 AM EDT: A user merged this pull request withGraphite.

Ok, now looks like

Host *.coder        ConnectTimeout=0        StrictHostKeyChecking=no        UserKnownHostsFile=/dev/null        LogLevel ERRORMatch host *.coder !exec "C:\Users\micha\repos\coder\coder.exe connect exists %h"        ProxyCommand C:\Users\micha\repos\coder\coder.exe --global-config C:\Users\micha\AppData\Roaming\coderv2 ssh --stdio --hostname-suffix coder %h

@spikecurtis works as expected now, thanks! It'd be nice to move those default ones (at least ConnectTimeout, StrictHostKeyChecking, and UserKnownHostsFile, but probably LogLevel too) to the match block though, since host key verification works well over Coder Connect (seems a shame to disable it) and the others are most likely related to proxying as well.

Honestly, we can probably retireStrictHostKeyChecking andUserKnownHostsFile settings entirely at some point soon, since we've implemented steps to ensure Agents use the same key on rebuilds.

Oh yeah, I thought it was still a requirement for the proxy command but it seems you are right, we can totally do that! I think theConnectTimeout andLogLevel ERROR are still required to be able to wait for a workspace to start and perhaps show output as well, but have not verified.