db-derby-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Shreyans Jain <shreyans2...@gmail.com>
Subject Re: Derby Database Corruption Issues
Date Wed, 12 Jul 2017 19:11:50 GMT
Also the same article displays " write cache techniques. Some hardware has
a default setting of 'write cache' set to enabled. This functionality will
interfere with Derby's expectation of what is written when to the disk, and
has shown to cause database corruption. "

Is it the same write cache that we can disable via OS on windows 10 from
Device Manager > Properties of Disk Drive > Policies tab or it has to be
disabled from somewhere in BIOS.

Shreyans Jain

On Thu, Jul 13, 2017 at 12:32 AM, Kristian Waagan <kristwaa@apache.org>

> Hi Shreyans,
> I believe this is still valid advice, but the details depend on your
> IO-system, your operating system and your server hardware.
> The issue here is that Derby is told the data has been written to stable
> storage, but then it has in fact not. If the machine crashes at this point
> data may be lost.
> For instance, if you are using a SAN there is typically a battery / UPS
> backed buffer (or other types of buffers that don't loose data in case of a
> power outage) on the "storage side", but for this to be useful you must
> ensure data is pushed through the IO driver and is not buffered locally by
> the operating system.
> Using asynchronous writes is a typical way of obtaining better [write]
> performance. In many cases the downside is a higher chance of data
> corruption.
> This may or may not be acceptable for your application.
> Regards,
> --
> Kristian
> Den ons. 12. jul. 2017, 20:04 skrev Shreyans Jain <shreyans2799@gmail.com
> >:
>> I was reading https://wiki.apache.org/db-derby/DatabaseCorruption and i
>> found prevention of corruption which states
>> "Switch off the machine's write caching" . now the article is written in
>> 2013. Is it still application to derby database corruption or now it has no
>> effect.
>> Regards,
>> Shreyans Jain

View raw message