[dm-devel] failed multipaths are never re-validated

[christophe.varoqui at free.fr]

> 
> I've not been able to test a controller failure; the HSVs I'm running
> against are in production for other systems. However, I should have access
> to a dedicated pair of HSV in a few weeks, so I'll be able to do the test
> I'll be able to test breaking the controller.
> 
In the mean time you can take much fun seing your IO load bump from one
controler to another by sending "sg_start -s 1 /dev/sd?" through inactive paths.
It only affects the specific LUN, so shouldn't be dangerous for your production.

<standard disclaimer about me not willing to assume your potential loss of data :>