upgrade to 10.3R ....

[Jon Radel]

On 9/28/16 3:50 PM, Lowell Gilbert wrote:
> "William A. Mahaffey III" <wam at hiwaay.net> writes:
> 
>> are many online URLs with dire warnings about using ZFS & non-ECC RAM
>> causing data corruption (this box is a commodity desktop w/ non-ECC
> 
>> 1. Are these concerns valid, i.e. founded in fact ?
> 
> Sure. When you get uncorrected bit flips, you can end up corrupting your
> data or even the filesystem. That isn't unique to ZFS, of course.
> 
> If the data integrity guarantees aren't the deciding reason you're using
> ZFS, then this isn't an issue. Some people use ZFS on laptops, for
> example, for other features rather than data integrity (most commonly,
> compression).

I'd argue that it isn't really an issue even if you use ZFS for data
integrity.  At least not in the way many of the statements on the
Internet seem to mean it....

I've read some where you get the sense that the message is:  ZFS with
non-ECC RAM means you are doomed.  Any other file system with non-ECC
RAM and everything is hunky-dory.  Well, no, not so much.  You write
garbage in RAM to disk and you're going to have garbage on disk.  Choice
of file system doesn't help with that.

I suspect that there is *some* danger of a higher chance of false
positives while scrubbing a ZFS file system using non-ECC RAM actually
inducing errors on disk instead of correcting them, and I'd find a
credible analysis of that a most interesting read, but really, given the
probability of bit error on today's very large hard drives, I'm more
worried about the long-term data stability of my disks than of my RAM.

That said, for the data I really care about, I use ZFS and ECC RAM....


-- 
--Jon Radel
jon at radel.com

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 3890 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://lists.freebsd.org/pipermail/freebsd-questions/attachments/20160928/14819686/attachment.bin>