Eucalyptus component registration process is manual

I think cluster registration is actually already handled (see tools/eucalyptus-cc.in:register_local_cloud in the Ubuntu branch), but some more similar work needs to be done on the other components.

Do you have any idea how we can tell in a script whether the Walrus and storage controller components are already registered? For the cluster controller, we check for the existence of /var/lib/eucalyptus/keys/cluster-pk.pem; but it isn't obvious to me how to do the same for the Walrus and SC.

I'm not sure that there is a good way to tell from a script whether they have been registered, but I can say that the registration command is idempotent. One can run 'euca_conf --register-walrus <ip>' many times, and only the first time will have any effect (consecutive runs will succeed, but the command is basically a NOP). Same for SC. Is this sufficient behavior?

What if it's already registered to another cloud controller, though?
Would --register-walrus <local-ip> forcibly re-register with a local
cloud? We wouldn't want that, I think.

We'd want some way to distinguish "failed to register because already
registered" from "failed to register because something went wrong". As
such I think it would be helpful to have a way to tell up-front whether
the component has already been registered or not.

I'd really appreciate it if somebody could eyeball this branch and let me know whether it looks sane:

  http://bazaar.launchpad.net/~cjwatson/eucalyptus/register-walrus-sc/revision/539

I think I have everything in place now, but in my testing the cloud controller was falling over in various ways when I tried to register components. I'm not at all convinced that that wasn't an artifact of my test environment, so I would appreciate somebody with a more competent environment giving this a run-through and seeing how it works.

This bug was fixed in the package eucalyptus - 1.6~bzr746-0ubuntu2

---------------
eucalyptus (1.6~bzr746-0ubuntu2) karmic; urgency=low

  [ Colin Watson ]
  * On initial cluster installation, allow authentication to the front-end
    using the cluster's SSH key (LP: #429087).
  * Update the cloud .jar name we look for when registering a cluster with a
    local cloud.
  * Check for IPv6 listeners as well as IPv4 when registering a cluster with
    a local cloud.
  * Actually kill cloud/Walrus/SC processes before entering a timeout loop
    to wait for them to die.
  * Automatically register Walrus and storage controllers with any local
    cloud (LP: #425922).
  * Discard standard output from euca_conf when registering components in
    init scripts.

  [ Soren Hansen ]
  * Add configuration for mpm_event.

 -- Colin Watson <email address hidden> Tue, 15 Sep 2009 22:44:25 +0100