knife ssh fails wholesale when one or more hosts are unavailable

Please consider this change to knife/ssh.rb :

1. diff /usr/lib/ruby/gems/1.8/gems/chef-0.9.6/lib/chef/knife/ssh.rb ssh.rb
   60c60
   < @session ||= Net::SSH::Multi.start(:concurrent_connections => config[:concurrency])
   —
   > @session ||= Net::SSH::Multi.start(:concurrent_connections => config[:concurrency], :on_error => :warn)

Perhaps a better fix is to report the error.
http://github.com/avishai-ish-shalom/chef/tree/CHEF-1137

setting fix version for 0.9.10

Merged to master: http://github.com/opscode/chef/commit/6a7ebf51f43a1db94904061687dba6d9982de637

I tweaked the error message a little bit so it will include the error class, which can be pretty helpful diagnosing a problem (for example Net::SSH::AuthenticationFailed if your username/password/ssh key is/are wrong).

Not sure this is fixed. If I have a node down in the search I get:

$knife ssh "role:base" "hostname"

System/Library/Frameworks/Ruby.framework/Versions/1.8/usr/lib/ruby/1.8/timeout.rb:60:in `open': execution expired (Timeout::Error)
from /Library/Ruby/Gems/1.8/gems/net-ssh-multi-1.0.1/lib/net/ssh/multi/session_actions.rb:37:in `join'
from /Library/Ruby/Gems/1.8/gems/net-ssh-multi-1.0.1/lib/net/ssh/multi/session_actions.rb:37:in `sessions'
from /Library/Ruby/Gems/1.8/gems/net-ssh-multi-1.0.1/lib/net/ssh/multi/session_actions.rb:37:in `each'
from /Library/Ruby/Gems/1.8/gems/net-ssh-multi-1.0.1/lib/net/ssh/multi/session_actions.rb:37:in `sessions'
from /Library/Ruby/Gems/1.8/gems/net-ssh-multi-1.0.1/lib/net/ssh/multi/session_actions.rb:81:in `open_channel'
from /usr/local/Cellar/gems/1.8/gems/chef-0.9.12/lib/chef/knife/ssh.rb:143:in `ssh_command'
from /usr/local/Cellar/gems/1.8/gems/chef-0.9.12/lib/chef/knife/ssh.rb:278:in `run'
from /usr/local/Cellar/gems/1.8/gems/chef-0.9.12/lib/chef/knife.rb:127:in `run'
from /usr/local/Cellar/gems/1.8/gems/chef-0.9.12/lib/chef/application/knife.rb:121:in `run'
from /usr/local/Cellar/gems/1.8/gems/chef-0.9.12/bin/knife:25
from /usr/local/bin/knife:19:in `load'
from /usr/local/bin/knife:19

$ knife -v
Chef: 0.9.12

Dan I observed this behavior with Brian...might be a regression.

If I remove this from my .ssh/config, it works

ConnectTimeout 5

However, it takes a long time to timeout.

Could not reproduce.

[514]avishai@gregol:chef$ sudo iptables -I OUTPUT -d monitor.fewbytes.com -p tcp --dport 22 -j DROP
[515]avishai@gregol:chef$ time knife ssh 'name:monitor*' hostname -x root
WARN: Failed to connect to node[monitor.fewbytes.com] -- Timeout::Error: execution expired

real	0m6.909s
user	0m1.190s
sys	0m0.170s
[516]avishai@gregol:chef$ knife -v
Chef: 0.9.12

CHEF-2300 bumped net-ssh-multi to 1.1, which may resolve this issue with a newer upstream.

My syntax for ConnectTimeout for 'knife ssh' in my '~/.ssh/config' file:

Host *
  ConnectTimeout 5

This is also broken for me on Chef 0.10.10, even after updating to net-ssh-multi 1.1. For example:

$ knife ssh -x me -C 10 -G gatewayhost.my.domain 'os:linux' 'echo foo'
host1.my.domain     foo 
host2.my.domain     foo 
host3.my.domain     foo 
host4.my.domain     foo 
host5.my.domain     foo 
WARNING: Failed to connect to node[host6.my.domain] -- Net::SSH::Disconnect: connection closed by remote host

After this warning, knife "hangs", eg nothing further happens and I have to ctrl-C to terminate knife.