Servers sometimes go pear-shaped. It's a fact of life – perhaps not so much with Windows 2000 as it was with NT 4, but it still happens. Emergency Repair for Windows (formerly RoboER) from Itheon (formerly Heroix) provides a means for system managers to connect a server that's in a state of limbo and either start and stop processes in order to resurrect it, or at least reboot it gracefully.
The nature of multi-user systems is that even though the system itself is running, processes can run away with the CPU, and/or take up all the available memory, and make the whole thing run slowly. When this happens, it's often the case that although the thing is limping along, you can't log in from the Windows console because whatever's caning the machine is running at a higher priority than the Windows GUI itself.
ERW provides a command-line interface that allows you to perform key maintenance tasks without the need for the Windows GUI. Because it runs at a very high priority, it doesn't suffer from the performance problems in the event that the CPU is maxed out, and so it's a useful back door into the low-level Windows configuration. Note at this point that it's more than just a DOS window (although one of the things you can do via ERW is execute DOS commands).
The application sits on the server and listens for connections on either a network interface or a serial port (or both). When you connect, it prompts for a username and password, and assuming you're authorised, it then presents you with a command line prompt. Incidentally, there are two methods of login – either via the Windows user database or via a built-in "back door" password that you can use even if (as often used to be the case with NT 4) the SAM lookup mechanism is refusing to work correctly. Of course, you need to be careful to change the back-door password frequently if you use it, and to limit the IP addresses from which it allows back-door connections (you can do this from the ERW control panel applet), but this is just common sense.
At the command line prompt you can do a number of things. First, you can watch the system events as they flash past, since ERW can be told to replicate event log entries to the console (you can also apply filters that reduce or eliminate what you see – there's nothing worse than trying to fix a problem amid zillions of error alerts). Next, you can choose to use a DOS prompt, which enables you to do anything that you'd do in a DOS box from the system console.
The important bits, though, are the commands that ERW gives you that wouldn't normally be available via a command line. First, you can either restart or shutdown the server – the last resort, but it's an option. Next, you can list the running processes and/or threads, to get an idea of what's run away with all the system resources. Next on the list is the SERVICE command, which allows you to start and stop services, list the dependencies of a service, see what's running, and even change the startup characteristics of any given service; basically it replicates the function of the Services control panel on an emergency command line.
Then you have the process handling function, which allows you to kill any given process, by name or process ID. This is, in our experience, the function that rescues you the most often, because once you've found which process is causing the problem, you can do a PROCESS /KILL on it and then perhaps use the SERVICE command to disable the auto-startup of that item.
The final useful item is a command-line REGEDIT – though admittedly it's not something you'd use by choice because registry keys are usually long and cryptic and it's much nicer using the Windows GUI-based REGEDIT application. But if you really need to hack a registry key, ERW lets you do it.
Alongside the fact that it does everything you might need for a wobbly server, ERW's other key strength is that because it's a command-line utility, it needs next to no bandwidth. Forget your Remote Desktop Connection or VNC rubbish – if you're in the middle of nowhere on a GSM cellphone and your monitoring system pages you to say the server's hosed, what would you rather have – a hideously slow GUI or a command-line?
I really like ERW. In fact I've used it in real life on a remotely located installation of nine NT 4 servers, where it saved literally thousands of pounds in person and travel time when things went a bit awry. In my opinion, anyone who runs Windows servers should also run ERW – particularly if you spend any time managing the servers from afar over low-bandwidth links.
When you're looking for network and system management software, you should be looking to build a "toolbox" of products – you're unlikely to get a single suite that does everything you need. So tools like this one would be just one part of a larger collection of products.