power failure causes files loss

[Eric Sandeen]

Hans-Peter Jansen wrote:
> On Saturday 24 April 2010, 00:25:41 Stan Hoeppner wrote:
>> adamantzj put forth on 4/23/2010 6:28 AM:
>>> Hello,
>>>
>>>      I’m writing data on a server in my company, but power cut
>>> suddenly. I had to restart the machine, then I found that many files in
>>> the working directory had been lost. I made some tries with xfs_repair
>>> and it didn't repair my corrupted file.
>>>
>>> What are the reasons caused this problem? I would like to know is what
>>> can I do to resolve this problem? Is XFS able to recover my file with
>>> its right content, at least a consistent content?
>>>
>>>      I'm running a 2.6.24 kernel with xfsprogs-3.1.1, and the operating
>>> system is RedHat Enterprise Linux 5 Server (64-bit).
>>>
>>>      Do you know if this problem has often been encountered, there is a
>>> way to proceed?
>>>
>>> Thank you !
>>>
>>> Adamant
>> XFS isn't the cause of your problem.  Running without a proper UPS is the
>> cause of your problem.  Fix the cause of your problem by purchasing a
>> properly sized and instrumented/monitored UPS.  YOU CANNOT rely on your
>> OS or filesystem driver to prevent data loss due to power loss, or to
>> properly recover from such unexpected power loss if/when it occurs.  This
>> is why you absolutely need to eliminate the possibility of unexpected
>> power loss.  This is the sole purpose for the existence of companies such
>> as APC, Liebert, TrippLite and others.
>>
>> I'd have thought by 2010 that the entire IT cadre worldwide would already
>> be educated on proper power backup planning and procedures.  I guess not.
>>
>> Acquire a proper UPS and configure it properly, specifically meaning
>> monitoring and alerts which will allow for performing a proper shutdown
>> during an extended outage, and your problems will disappear.
> 
> While you're right for the general case of course, there are many other 
> failure conditions, that an UPS will not prevent, eg. kernel crashes. 
> Therefore there is an vivid interest, that any storage based filesystems  
> should stay consistent _any_ time (and usually filesystem developers spend 
> an considerable amount of their time on fixing such problems, if they 
> arise). 
> 
> Adamant, the people reading here are rarely trained in telepathic   
> capabilities. If you need more specific answers, you need to provide more 
> specific details, eg. mount options, error logs, etc.
> 
> Pete

See also:

"Eat My Data: How Everybody Gets File IO Wrong"
http://www.flamingspork.com/talks/

:)

-Eric