Nov 242016
 

Broken Windows logo[1] I know what I should do if a system service on Microsoft Windows starts crashing of course; Fixing it is the way to go! But sometimes you simply can’t, because the component causing a certain instability can’t be swapped out or updated. Now Windows services do have a mechanism for monitoring and restarting a service upon failure, but it seems that only works if the system gets an actual error code back from the service upon termination. But it doesn’t seem to work (at least for me) if the service just dies abnormally. Windows recognizes the service has stopped somehow of course, but the restart procedure just doesn’t kick in.

So I thought I’d do it myself, programmatically. And it’s actually pretty easy. I solved this with VBScript, Windows Batch and Mark Russinovichs’ pslist plus grep. So the prerequisites are:

  • Microsoft Windows (well, huh..)
  • MS Windows Script(ing) Host / VBScript, Windows should come with this preinstalled since Windows 2000.
  • [pslist]
  • [grep][src] (grep is optional, I used GNU grep 2.5.4 in this case, licensed under the [GPLv3+])

Make sure the pstools and grep are within your %PATH%, so Windows can find those .exe files. If you don’t want to use grep, you can also use Microsofts’ own find command, if your version of Windows has it.

I divided this into two small scripts. Since the main part is Batch, it might be problematic if you run it at very short intervals, checking for the services’ status, because you get a command window popping up on the desktop. Since most users wouldn’t want that, another script acts as a launcher, hiding the cmd.exe window so it’s run fully in the background without disturbing any potential users or administrators. The launcher looks like this, in my case it’s meant to watch over an Apache web server:

  1. Set WshShell = CreateObject("WScript.Shell")
  2. WshShell.Run chr(34) & "C:\Server\Scripts\monitor-httpd.bat" & Chr(34), 0
  3. Set WshShell = Nothing

And that script C:\Server\Scripts\monitor-httpd.bat we’re launching looks like this:

  1. @ECHO OFF
  2. FOR /F "tokens=* delims= usebackq" %%I IN (`pslist ^| grep httpd`) DO SET HTTPDSTATUS=%%I
  3. IF NOT DEFINED HTTPDSTATUS (net start "Apache2.2") ELSE (SET HTTPDSTATUS=)

A version relying on Microsoft find instead of GNU grep could look like this:

  1. @ECHO OFF
  2. FOR /F "tokens=* delims= usebackq" %%I IN (`pslist ^| find /I "httpd"`) DO SET HTTPDSTATUS=%%I
  3. IF NOT DEFINED HTTPDSTATUS (net start "Apache2.2") ELSE (SET HTTPDSTATUS=)

To get a services’ exact name, just launch services.msc from Start \ Run or run the command net start on a cmd terminal.

As you can see, this greps “httpd” from the process list and pushes its output into %%I and finally into %HTTPDSTATUS%. We have to use a FOR /F for that, as Windows has no way of pushing command outputs from subshells into shell variables like UNIX has (like e.g. var=`command` or var=$(command)). Then we check for the status of that variable. If it’s not defined, then the process http.exe was nowhere to be found! In that case we restart the associated system service (needs proper permissions!). If the variable is defined, we do nothing but unsetting it, since we can assume the service is operating normally. Or at the very least it’s running. ;)

You can automate that by using the Windows task scheduler:

Scheduling an Apache web server "watchdog"

Scheduling an Apache web server “watchdog” (German Windows)

Create a Schedule to your liking and you’re done! If you can afford the affected service to be down for 5 minutes and no longer, just run it every 4 minutes or so.

The solution shown above can easily be adapted to monitor and restart any Windows service you have, as long as the service isn’t fundamentally broken so that it wouldn’t even start up anymore. Also, you can do a lot more, like sending notification eMails with a command line mailer like [blat] when crashes do occur. Of course, this is only useful for services that crash rarely. If it dies every few minutes, you should reaaally fix it instead of just pushing the restart button all the time… ;)

And that’s that!

[1] © Mar.0007. Original Version for desktopwallpapers4.me.