mismatch_cnt, random bitflips, silent corruption(?), mdadm/sw raid[156]

[Justin Piszcz]

On Thu, 25 Dec 2008, David Lethe wrote:

>> -----Original Message-----
>> From: linux-raid-owner at vger.kernel.org [mailto:linux-raid-
>>
>> Other options?
>>
>> How do others maintain data integrity?  Just not worry about it until
>> you have
>> to, rely on backups.. or?
>>
>> Justin.
>
> 4GB files using gpg and tar in the '90s?
Sorry, to clarify they were ~650-700MiB tars but combined to around a ~4GiB
file later on (from the late 90s), CD's were cheap and yeah 2GiB limit.  But
later on to consolidate I moved them to 4GiB DVDs and thus tarred them together
and then gpg on top of that.

> I know gpg had 2GB file-related bugs as late as 2005 that caused corruption, and
> there was a heck of a lot of 2GB-related bugs in 2.2 and 2.4 kernels which you
> must have been running back then.  You are also using later versions of these
> programs on the new systems, and I'd be willing to bet they compound the problem by
> assuming there was no corruption to begin with.
Once I restored the data from DVDs, I was able to restore *all* data 
successfully.

>
> I use:
> - gzipped tar archives, but I gzip the individual files, rather than the tarball. That way
> any compression-related bugs are limited to a single file. I copy them to DVDs.
That works as well, but are they your regular files gzipped, no encryption?

>
> - For online/nearline, I now use a ZFS, but on a native Solaris system that functions as my primary
> NFS/CIFS/iSCSI server with and a ZFS software-RAID based file system.   I am profoundly impressed with
> it, and when they release the deduplication enhancement for ZFS, I'll adopt it and won't have to buy
> any more DVDs, except for offsite archiving purposes.
Wow, I did not know ZFS had plans for de-dupe!! I will have to look into this,
thanks for the info.

Justin.