destroying containers from code doesn't sync with lxc-ls

[bararchy]

I created 10 containers, then, destroyed them all (muahaha).
Looking at lxc-ls and ifconfig I can see that the containers are still "running" but trying to execute lxc-destroy --name *** gives back an error

f87efd19-4594-407d-9b18-712273144675  RUNNING  10.0.3.130  -     -       NO         

root@Sphere:/var/lib/lxc# lxc-destroy --name f87efd19-4594-407d-9b18-712273144675
Container is not defined

looking at /var/lib/lxc shows an empty dir.

So it seems I got "phantom" containers ?

[andrenth]

Hi Bar

Can you share the script you ran that resulted in this mess? :)

Em 14 de jan de 2016, às 11:36, Bar Hofesh notifications@github.com escreveu:

I created 10 containers, then, destroyed them all (muahaha).
Looking at lxc-ls and ifconfig I can see that the containers are still "running" but trying to execute lxc-destroy --name *** gives back an error

f87efd19-4594-407d-9b18-712273144675 RUNNING 10.0.3.130 - - NO

root@Sphere:/var/lib/lxc# lxc-destroy --name f87efd19-4594-407d-9b18-712273144675
Container is not defined
looking at /var/lib/lxc shows an empty dir.

So it seems I got "phantom" containers ?

—
Reply to this email directly or view it on GitHub.

[bararchy]

Sure,

To create I use this:

def create_container
    container_name = SecureRandom.uuid
    c = LXC::Container.new(container_name)
    begin
      c.create('ubuntu')
      c.start
      SYSLOGGER.info("Pool: Starting container: #{c.name}")
    rescue => e
      SYSLOGGER.error("Pool: Error starting container #{c.name}: #{e}")
    end
    return c if c
  end

To destroy I use this:

def destroy_container(c)
    container_name = c.name
    try = 0
    begin
      while c.running?
        SYSLOGGER.debug("Container: Shutting down container #{c.name}...")
        c.shutdown(-1)
        sleep 0.1 while c.running?
        SYSLOGGER.debug("Container: Stopping container #{c.name}...")
        c.stop
        c.wait(:stopped, 5)
        SYSLOGGER.debug("Container: Destroying container #{c.name}...")
        c.destroy
        SYSLOGGER.debug("Container: Clearing config for container #{c.name}...")
        c.clear_config
      end
      SYSLOGGER.info("Container: Container #{container_name} closed")
    rescue => e
      SYSLOGGER.error("Container: Error stopping Container #{container_name}: #{e}")
      sleep 2
      try += 1
      if try < 3
        retry
      else
        @running = true
      end
    end
  end

So i basically did something like

container_array = []
10.times do 
  container_array << create_container
end

container_array.each do |c|
  destroy_container(c)
end

[bararchy]

@andrenth , did you had time to check this issue ? Can I help somehow ?

[ranjib]

what distro you are using? Can you run the script and print lxcpath as part of it?

puts LXC.global_config_item('lxc.lxcpath')

fwiw, I tried your script (with unprivileged containers) and it worked fine for me.

[bararchy]

Hi @ranjib , I'm running:

DISTRIB_ID=Ubuntu
DISTRIB_RELEASE=16.04
DISTRIB_CODENAME=xenial
DISTRIB_DESCRIPTION="Ubuntu Xenial Xerus (development branch)"

puts LXC.global_config_item('lxc.lxcpath')
=> /var/lib/lxc

I'm running with Privileged containers.

Ruby:

ruby --version
ruby 2.2.1p85 (2015-02-26 revision 49769) [x86_64-linux]