[time-nuts] 5370 processor boards available

[cheater00 .]

Hi Brian,
it's less about caching pending writes to file handles that are
waiting inside the system. That's most likely to work well within a
short period of time.

The issue is if you're using the system to its full extent. Among
others you'd like to write applications for the BBB that enhance the
5370's capabilities. You don't want to bother with hard real time,
compilation, memory layout, etc, all of them problems each of which
will make your development time 10x longer and your fun time 20x
shorter. So instead you use something like Python or Java. Then, you
also don't really bother to ensure that your program can be
interrupted at any time, since that's another level of complexity.
Coroutines? Continuation passing style? Reactive programming? Why
bother? Then you make a few other decisions that make your program
easier to make and more difficult to shut down immediately in a clean
manner.

So let's say it's running and it is collecting data which is crucial
to you. Maybe you again didn't bother to make sure that if the program
crashes it will not corrupt its data.

So now your BBB gets the power loss interrupt, sends kill -15 to all
processes and most of them exit. Your program doesn't so likely Linux
tries to send signals 2, then 1, then 9 at which point your program
will certainly crash (sig 9 means "crash now"). So your program will
likely eat all the data.

Note that all of the shortcuts I mentioned above are fairly reasonable
defaults that save days of work per project for no improvement in
reliability, since, after all, a suitable battery only costs $5.

Cheers,
D.

On Thu, Feb 27, 2014 at 7:47 PM, Brian Lloyd <brian at lloyd.com> wrote:
> It is interesting to watch this thread. The most interesting thing is the
> fact that no one has asked or answered the basic question: is there a
> problem? There are a lot of solutions for what may be a complete non-issue.
>
> Given that Linux and BSD are reliable and stable on standard PC platforms
> connected to mains power without a UPS, that suggests to me that there is
> likely no issue. Sure, back in the bad ol' days you had to sync;sync;sync
> the filesystem prior to shutdown and then, if there had been a power fail,
> fsck it on power up. But those days have been gone for a LONG time.
>
> The modern filesystems reduce the exposure to the possibility of file
> system corruption to a tiny probability, and then the system further
> reduces that by providing a power-fail detection system that allows the
> critical pending writes to be flushed prior to final power-fail, thus
> leaving the FS in a completely deterministic state.
>
> I suppose that on the night of a full moon when the lightning flashes and
> the dog barks, conditions might be right to corrupt the file system, but
> then, that could happen when the battery goes dead or catches fire too. ;-)
>
> It seems to me that the bard put it pretty well, "... much ado about
> nothing."
>
> --
> Brian Lloyd, WB6RQN/J79BPL
> 706 Flightline Drive
> Spring Branch, TX 78070
> brian at lloyd.com
> +1.916.877.5067
> _______________________________________________
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.