[Discuss] SysVinit vs. systemd

[Bill Ricker]

On Wed, Sep 17, 2014 at 1:10 PM, Richard Pieri <richard.pieri at gmail.com> wrote:
> This may fail if you're using Multipath I/O.


Heh, Multipath i/o can break almost anything, but particularly when in
error cases.
Subtle interactions between bugs in different layers that aren't
*singly* a problem but ... are really hard to test for, or replicate.

I've had more failures made worse by Multipath than saved by it
(although not on Linux, and not a statistically significant quantity:
merely plural anecdotes so far; but both with OS-vendor MPio and
Storage-vendor MPio *sigh*).
    (E.g., in one case, OS & MPio bugs interacted when a card or
controller was bad - had we not had MPio, it would have been simple HW
crash to route away from and repair; with MP it behaved weirdly,
badly, and required debugging, as MPio masked the problem and
expressed it elsewhere.)

So MPio can be (mostly) good for a critical DB server's data storage,
i might use it again in similar max-uptime situation, but not as
ubiquitously; for other things, other more virtual forms of disk
access may be more resilient in reality. (And they're coming for our
DBs too.)

-- 
Bill Ricker
bill.n1vux at gmail.com
https://www.linkedin.com/in/n1vux