On Wed, 22 Oct 2014 14:31:02 -0400, Barry Shein <bzs@world.std.com> wrote:
Perhaps you don't remember the days when an fsck was basically mandatory and could take 15-20 minutes on a large disk.
Journaling has all but done away with fsck. You'd have to go *way* back to have systems that ran a full fsck on every boot -- and in my experience, you absolutely wanted that fsck. I've used xfs for over a decade. It doesn't even have an fsck (xfs_check and xfs_repair, yes, but NO system will ever call them. And as a rule, never needs to.)
And you whisk all that away with "it's not really clear to me that 'reboots in seconds' is a think to be optimized"????
You're arguing the difference between optimizing a 15min boot into a 5min boot, vs a 15sec "boot" into a 14sec "boot". The former is an actual optimization allowing subsystems to start in parallel, in ways that do not introduce delay. (eg. sendmail startup, used to be the #1 slow down on solaris) The latter is pure nonsense, with "boot" being measured as when login pops up -- which is NOT when all the subsystems have actually completely started.
To me that's like saying it's not important to try to design so one can recover from a network outage in seconds.
Your efforts are better spent avoiding an outage in the first place. If outages are common enough to be something that needs to be "sped up", then you've already failed.