[Info-vax] Re; Spiralog, RMS Journaling (was Re: FREESPADRIFT)

[lawrencedo99 at gmail.com]

On Sunday, June 19, 2016 at 6:35:05 AM UTC+12, Kerry Main wrote:

> With OpenVMS and its default write-through strategies, data integrity 
> has always been the primary driver. 

I fully accept that. And I, too, deplore vendor strategies that prioritize good-looking benchmarks over data integrity. Particularly insidious are hard drives with on-board cache, where a request to “make sure you really, really write this data to disk” can be met with “OK, I really, really have done it” right down to the SATA interface level, only to have the data still sitting in some controller cache buffer, not really on the disk at all, so it could still be lost on the next power failure.

Filesystem journalling is a way of improving your confidence in the integrity of the volume structure across crashes. So instead of doing a fsck (MOUNT/VERIFY?) on every reboot, you can defer that to, say, once every six months (I think that’s what the Linux systems usually do).