EmergencyShutdown

Describes how the local site administrator handles emergency shutdown requests.

How do I find the list of active slices on my Aggregate?

To see the slice list, go to Emulab web interface Administration drop down menu (in red-dot mode), and click on ProtoGeni Slices. There you will see a list of all active slices at your aggregate.

Basic Shutdown

Instageni shutdown operates at the slice level; all resources at a given aggregate that are associated with the slice in question, are terminated, and the nodes restored to a clean condition. It is not possible to emergency stop a single sub component of a slice (ie: individual node). This might be supported in the future though.

The primary interfaces for shutdown are through administrator scripts that are run on the local boss node. There are two forms of shutdown. First, the site administrator can easily terminate a slice without raising the global panic level to red, by running the following script.

 boss> wap cleanupslice -m slice_urn

All resources are released, rebooted and the disks restored to a completely clean state.

If however, the local administrator wants to signal a global shutdown of the slice terminating the slice on all aggregates (and the slice is named by the local Slice Authority, not a foreign authority), a different script is used:

 boss> wap shutdownslice slice_urn

This will result in an XMLRPC message to the ProtoGeni clearinghouse in Utah, which in turn will issue an XMLRPC message to all aggregates to shutdown the slice on their aggregate. As mentioned above; this is a total termination of all resources on each aggregate.

How do I confirm that a slice has been terminated?

To confirm that a slice has been terminated, go to Emulab web interface Administration drop down menu (in red-dot mode), and click on ProtoGeni Slices. There you will see a list of all active slices at your aggregate, and thus can confirm that the slice has indeed been terminated.

How do I map a resource to a slice?

If you do not know the slice urn, but you do have a sliver urn, or an IP, or maybe a MAC address, you can use the maptoslice script on boss to find out what slice a particular resource is part of. This information in also available in the Emulab web interface in Administration drop down menu. To search for a slice using the command line:

 boss> wap maptoslice urn:publicid:IDN+emulab.net+node+pc488
 urn:publicid:IDN+pgeni.gpolab.bbn.com+slice+d4

Other forms of the command:

 boss> wap maptoslice urn:publicid:IDN+emulab.net+sliver+123
 boss> wap maptoslice urn:publicid:IDN+emulab.net+user+stoller
 boss> wap maptoslice 155.98.33.38  ; a control IP address
 boss> wap maptoslice 00AABBCCDDEE  ; a control or data mac address

Future Work:

While we cannot remove a sliver from a slice, we should be able to shut that sliver down, perhaps by powering the node off. This is a little less drastic then terminating the entire slice on the aggregate.

Emulab has the ability to isolate an experiment without completely terminating it, by putting the control network interfaces in a black hole vlan. From that point on, the only way to interact with the nodes is via the console, but this allows us look inside if we need too.