[Samba] File Corruption

[Scott Sipe]

Jeremy Allison wrote:
> On Fri, Aug 01, 2003 at 11:27:44PM -0400, Scott Sipe wrote:
> 
>>As another example, I had a mozilla installer sitting in a directory on 
>>the share--no one but me would have touched it.  Last used ~1 month ago. 
>>      Tried it today and it was corrupt.
>>
>> [snip]
>>
>>Does anyone have a clue what the problem might be, or what I can test? 
>>The problem with this is that it IS a sporadic problem, but it's 
>>happened to me enough times that I'm sure it's real.
> 
> 
> This will need more information to track down I'm afraid. What makes
> me very suspicious if the mozilla installer example. Can you try installing
> an audit vfs to track who wrote what to a file ? That way you can
> track if Samba has accessed the file at all or if it is another process
> on the machine.
> 
> Jeremy.

Ok, I added the auditing module to my main share.  As a test I
redownloaded the mozilla installer and put it in it's own directory on 
the samba share.  I took a md5 immediately after downloading it.  I then 
tried installing to several computers directly from the share, copied 
the file to client PC's and then back to the samba server, etc--it 
remained good.  Then I added a cronjob that mailed me the md5 of the 
mozilla installer file every hour, and went home.  The first email 
(mailed like 1 minute before I left work) was the correct md5, 1 hour 
later (I was not doing anything, nor was anyone else for that hour) the 
file is corrupt.  The audit log shows no access to that file from samba.

So I guess the only conclusion is that it's not Samba's fault?  The
thing I still don't understand is, if something is wrong with the system
(disk corruption?  something else weird?) then why I have had no other
problems?  This server has been running for maybe...2-3 years, is
currently at 50 days uptime, has had the world built many times, built
many ports, and I've never seen this corruption anywhere but on the
samba share.  I've set up some more tests to see if there really is 
corruption elsewhere, and to try to find out when/why it happens.

Thanks for helping me get at least a little bit farther in my diagnostics :)

thanks,
Scott