em-ssh-shell fails to start shell without error

[darkskiez]

Not sure whats happening yet, but my block inside EM::Ssh::Shell.new never gets evaluated, and no errors are raised

I can simulate this easily by connecting 50 times concurrently

[gruis]

Can you post some example code please?

Sent from my iPhone

On 2012/04/24, at 19:18, Mark [email protected] wrote:

Not sure whats happening yet, but my block inside EM::Ssh::Shell.new never gets evaluated, and no errors are raised

I can simulate this easily by connecting 50 times concurrently

---

Reply to this email directly or view it on GitHub:
#5

[darkskiez]

EM.run {
  EM.error_handler do |e| 
    puts " **** ERROR #{e.message}"
    puts e.backtrace.join "\n" 
  end
  started=0

  (1..30).each do
    EM::Ssh::Shell.new("localhost", "emtest", "emdemo") do |shell|
      started+=1    
      shell.on(:error) do |e|
        puts "Shell error :( #{e}"
      end
      shell.on(:closed) do |e|
        puts "Shell close :( #{e}"
      end
    end
  end

  EventMachine::PeriodicTimer.new(1) do
    puts "We have #{started} connections"
  end
}

On my machine stared goes to 10, but never get any exceptions or errors.

[darkskiez]

I found the reason for the 10 limit is the sshd_config default for MaxStartups is 10
but it doesn't explain the lack of errors.

[seryl]

I'd put my money on this actually being the thread pool in EM (I believe it defaults to 20)...

Also, opening 30+ requests is quite inefficient, I'd suggest looking into using EM's work and request pooling:

https://github.com/eventmachine/eventmachine/blob/master/lib/em/pool.rb#L9-38

[seryl]

root 2106 1 0 Aug13 ? 00:00:28 /usr/sbin/sshd
root 19326 2106 0 11:02 ? 00:00:00 sshd: root@pts/0
root 24137 2106 0 18:01 ? 00:00:00 sshd: root@pts/1
root 24154 2106 0 18:01 ? 00:00:00 sshd: root@pts/2
root 24155 2106 0 18:01 ? 00:00:00 sshd: root@pts/3
root 24156 2106 0 18:01 ? 00:00:00 sshd: root@pts/4
root 24157 2106 0 18:01 ? 00:00:00 sshd: root@pts/5
root 24158 2106 0 18:01 ? 00:00:00 sshd: root@pts/6
root 24159 2106 0 18:01 ? 00:00:00 sshd: root@pts/7
root 24160 2106 0 18:01 ? 00:00:00 sshd: root@pts/9
root 24161 2106 0 18:01 ? 00:00:00 sshd: root@pts/8
root 24162 2106 0 18:01 ? 00:00:00 sshd: root@pts/10
root 24163 2106 0 18:01 ? 00:00:00 sshd: root@pts/11
root 24305 24139 0 18:01 pts/1 00:00:00 grep ssh
[root ~]# ps -ef | grep ssh
root 2106 1 0 Aug13 ? 00:00:28 /usr/sbin/sshd
root 19326 2106 0 11:02 ? 00:00:00 sshd: root@pts/0
root 24137 2106 0 18:01 ? 00:00:00 sshd: root@pts/1
root 24307 24139 0 18:01 pts/1 00:00:00 grep ssh

I was incorrect. Tested the pool, not the case.

It's the max_startups as described above.

However, this still should be handled in a work queue, 10 simultaneous connections is plenty.

[kt97679]

Folks, do you still see this issue?