System changes

[drgrussell]

You should not really notice any of the changes made over the last few days. These include:

• Firewall changes. "nmap" was hanging up the server on which the virtual machine running nmap was executing. The firewall has been updated to handle nmap queries with REJECT, and to do this on the server itself.
• Race condition. It was possible to leave a virtual machine running while your browser was shut down. The conditions for this were exotic, but possible. Now unqueueing and requeueing takes up to 30 seconds, and the race condition possibility is exceptionally unlikely (e.g. practically impossible). Dequeueing and requeueing may result in an error message asking you to retry in 30 seconds.
• Getting onto a different server. If you dont like your server then help may be at hand. Booting and asking for a fresh image may put you onto a different server if you current server is strangely overloaded. Still checking the code for this one, but its mostly there.

[drgrussell]

More changes:

There was a tidy up of some code to do with queueing, and this involved some changes to the format of the queue data and queue handling. Any problems with the queue handling let me know. Such problems are likely to only be present in termination handling (i.e. leaving the queue or getting kicked out). The only noticable impact is that the action log for a user should contain a (slightly) clearer indication of why they were thrown out of the queue.

In addition the watchdog process now monitors all machines using an independent protocol, and shuts things down which do not appear in the queue file. This should never have any impact on normal users but gives me another line of defence against hackers trying to run a machine improperly.