Re: PKGBase and Embedded Systems

> On 12 Aug 2025, at 14:55, Karl Denninger <karl@denninger.net> wrote:
> 
> […]
> Its not so much a "how cheap is the resource" problem (yes, that's gotten a lot cheaper over time) its a "the box MUST come back online after an unsolicited power event."
> That in turn means the physical volume cannot be open for write, particularly in the instance of a volume that co-mingles various metadata that is internal to the device itself (e.g. an SSD which does its own internal wear leveling and such.)  95% of the time IMHO isn't good enough.

Our critical stuff is all on secured power, but we did do a certain amount of ‘plug-pull’ testing with ZFS and SSDs on PCEngines boxes and Intel rackmount servers, and we didn’t manage to make it misbehave.

Even if the physical volume isn’t open for write, it could still get trashed if the controller misbehaves in response to a power glitch.

> -- 
> Karl Denninger
> karl@denninger.net
> The Market Ticker
> [S/MIME encrypted email preferred]


--
Bob Bishop       t: +44 (0)118 940 1243
rb@gid.co.uk     m: +44 (0)783 626 4518