Re: bhyve and controlled errors

From: Dave Cottlehuber <dch_at_skunkwerks.at>
Date: Sat, 02 May 2026 13:45:16 UTC

On Tue, 7 Apr 2026, at 10:32, Miroslav Lachman wrote:
> On 26/03/2026 15:56, Sean Eric Fagan wrote:
>> I’d asked Mark about this and he suggested I bring it up on the list:
>> 
>> Has anyone thought about implementing controls / tunables to bhyve to introduce errors? I am, most specifically right now, thinking about causing disk I/O errors, or having requests dropped by the "hardware," to test error handling.
>> 
>> Last time I looked at the bhyve source code, it scared me, so if someone else has looked at this, I’m love to know.
>
> I understand it would be nice to have it directly as a part of bhyve, 
> but if you need something right now, you can try "gnop".
> See man gnop(8):
>       Its main purpose is testing other GEOM classes, as it allows
>       forced provider removal and I/O error simulation with a given
>       probability.
>
> Or maybe someone with skill can reuse its code for bhyve feature.
>
> Kind regards
> Miroslav Lachman

Chuck Tuffli already did some work in this area:

https://papers.freebsd.org/2019/bsdcan/tuffli-frankensteins_disk_drive/
https://bhyvecon.org/bhyveconOttawa2019-Chuck.pdf

You should be able to find the related videos for these.

A+
Dave