Fault Management testing in VirtualBox

[Jondice]

I'd like to simulate various hardware faults to have a better idea of what to expect of my production systems. I think this would be useful for systems developers (so maybe they know the answers) as well as careful administrators.

For instance, I could test a fault ina guest OS by deleting a disk image without telling the OS about it, but how can i test similar faults while the guest OS is powered on?

If you can't currently do this, maybe a permissive testing mode would be useful for virtualbox.

[Perryg]

This has actually been asked before and it is really not practical due to the way things are done in a virtual environment as compared to a real environment. IMHO

[vbox4me2]

Maybe the word 'virtual' has the wrong meaning for you, in essence any action inside a VM should not effect the Host thereby doing such tests are meaningless.

[mpack]

in essence any action inside a VM should not effect the Host thereby doing such tests are meaningless.

I don't see any indication in the OPs message that he expects guest faults to afflict the host.

It seems to me that the OP wants to do hardware fault injection testing, to test how software (running in the VM) copes with simulated hardware failures. However deleting entire disk images still seems a little extreme - I'm not sure what credible real world fault this would simulate.

OTOH I agree that this isn't what VBox is designed for, nor should it be.

[vbox4me2]

I can imagine this possiblilty with a self compiled version, thats where you need to create the fault switch.

[Jondice]

in essence any action inside a VM should not effect the Host thereby doing such tests are meaningless.

I don't see any indication in the OPs message that he expects guest faults to afflict the host.
...
OTOH I agree that this isn't what VBox is designed for, nor should it be.

Correct, I'm not wanting to afflict the host in any way. Deleting a disk was just a none too subtle example that is easy to do.

I suspected that Sun might also be interested in this, since they are working on fault management in solaris. I'm not very familiar with this area myself, though it is interesting. For practical purposes, I just want to test out what other people have done, so that I know what to expect should my production system fail in some way.

I was already able to test more or less what I wanted by removing a disk image ... so that was basically good enough for me .