At 08:53 AM 8/9/2008, Bernd Walter wrote:
>On Sat, Aug 09, 2008 at 02:04:28PM +0200, Wim Vandeputte mailinglist
>only wrote:
> >
> > On Sat, Aug 09, 2008 at 12:53:23PM +0200, Bernd Walter wrote:
> > > You can easily get a CRC read error for a logical block far away from
> > > the block which was written during power fail.
> > > So a rw partition can harm a ro partition - at least for power failures,
> > > which I've seen quite often in practice with all vendors I'd used.
> >
> > This is one more reason why I always advise people to buy industrial
> > grade flash
> >
> > http://www.siliconsystems.com/technology/powerArmor.aspx
Thanks to all those who responded! It sounds like even consumer
grade flashes have a decent write life cycle, but I certainly didnt
know about the power failure issue and corruption. How likely is
this to happen, as some of the sites I plan to deploy to will have
periodic power cycles / power issues.
I will try a quick test perhaps by power cycling a soekris a few
hundred times while writing out to a file on a RW partition and see
if I get any data corruption... I wonder though how many times would
I need to do the test and how many different flashes would I need to
sample in order to get meaningful results.
Also, for FreeBSD, what is my best option for the file system ? In
the past I stuck with UFS without softupdates as I would periodically
get issues where I could no longer mount the partition RO after
remounting it RW to do a file update. When I do need to write to the
flash, is it better to leave it async or sync ? i.e. does mounting
async allow the OS to write more intelligently to the underlying
flash ? After I finish writing, I typically remount it RO.
---Mike
_______________________________________________
Soekris-tech mailing list
[email protected]
http://lists.soekris.com/mailman/listinfo/soekris-tech
