This is Tim Abbott's initial draft at our automatic remctl
configuration.

/etc/remctl/sipb-xen-auto/remctl-update.sh is the magic script.  Run
it with "all" as an argument, and it will update everything.

The inputs to this system are as follows:

/etc/remctl/sipb-xen-auto/acl/MACHINENAME

  This directory contains files named MACHINENAME for each machine.
These ACL files specify who is allowed to administer the machine.  You
can use entries that are Kerberos principles, or entries of the form

include /etc/remctl/sipb-xen-auto/moira-acl/sipb-xen

to include AFS groups in ACLs.  To add a new machine to the system,
you simply need to create /etc/remctl/sipb-xen-auto/acl/MACHINENAME
and then run

/etc/remctl/sipb-xen-auto/remctl-update.sh all

Everything else is autogenerated from that information.


Other files of interest:

/etc/remctl/sipb-xen-auto/auto-machine-list

  The list of machines that should have their remctl configuration
files generated from the template.  This is generated from
listing /etc/remctl/sipb-xen-auto/acl/*.
 
/etc/remctl/sipb-xen-auto/auto-moira-list

  The list of Athena AFS groups from which acl files should be
generated.  The ACL files are placed in
/etc/remctl/sipb-xen-auto/moira-acl/, and named GROUPNAME.  Ths list
is generated by parsing the ACL files in /etc/remctl/sipb-xen-auto/acl/.


This package also includes a crontab to run

/etc/remctl/sipb-xen-auto/remctl-update.sh all

every 15 minutes or so to keep our Moira mapping up to date.  One can
request an update of our Moira mapping for group X by running

/etc/remctl/sipb-xen-auto/remctl-update.sh moiragroup X

The web interface should probably run this when it adds a group.  We
may want to make this also available to users, but I've been lame.

This package includes a remctl interface available to anyone to invoke
the command:

/etc/remctl/sipb-xen-auto/remctl-update.sh all

using the following command from your favorite machine with remctl:

remctl black-mesa.mit.edu remctl-auto-update all

It requires no special permission to run; there is a potential DOS
issue here, but I don't think it is serious.

Thought should be put into how to ensure that the servers stay in sync.