Drobo

Not Drobo-specific, but data integrity related (md5 hashes)...

Since this is “general” discussion and related to data integrity I was wondering if anyone had any insight into this oddity I run into occasionally at work.

The setup involved here is:

  • Mac Pro
    -> G-SPEED esPro (mini-SAS)
    -> three DroboPro’s daisy-chained in FW800
  • Fujitsu Eternus backup server (volume mounted via smb)

Whenever a video is completed here, I copy it to both a DroboPro and the Fujitsu Eternus so I have two RAID-6 copies. I also run md5 hashes on them to verify they copied correctly.

On occasion, like this morning I’ll get a result like this:

video.mov
md5 (Mac Pro local drive): value 12
md5 (DroboPro): value 13
md5 (Fujitsu): value 13

So the md5 hashes don’t match after the file is copied, but they’re identical on both the DroboPro and Fujitsu. Then I re-copy them, and run md5 from Terminal again and they’re all the correct value (in this simplified case, 12).

Isn’t that odd? I can understand a copy going bad because of some network issues or something, but how can the two bad copies have identical md5 hashes to one another?

Also, does anyone know of a good OS X program with a GUI to run batch checksums like this? There was a program I used years ago called “zsCompare,” but that company seems to be long out of business and there’s no way to get their software anymore.

I used to have a similar problem back in the Windows 2000 days. Somehow a bad copy of the file would get cached in the network somehow.

Repeated copies of the same file would produce the exact same bad output, even though the original source file was OK.

In the Win 2000 case, it was fixed by disabling OpLocks. Not sure if Mac OS has similar, and I haven’t seen the issue past Win XP…

Do you have a hardware RAID controller on the Mac? I’ve had hardware RAID with memory bit errors before too. :frowning:

Sorry can’t be of more help, but it sounds like a similar issue of some kind of cache gone bad.