You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lucene.apache.org by Jason Rutherglen <ja...@gmail.com> on 2008/11/19 18:44:45 UTC
Re: Option to fsync files
http://www.h2database.com/html/advanced.html#durability_problems
http://hardware.slashdot.org/article.pl?sid=05/05/13/0529252
On Wed, Nov 19, 2008 at 9:23 AM, Mark Miller <ma...@gmail.com> wrote:
> /A look at the slashdot article seems to indicate the OS may have been at
> fault in his case (an update he posted in response to some slashdot flames).
> Thats interesting. Wasnt there the last time this topic came up. Can't
> remember what java db had the other info...but looking...
>
> /
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: java-dev-unsubscribe@lucene.apache.org
> For additional commands, e-mail: java-dev-help@lucene.apache.org
>
>
Re: Option to fsync files
Posted by robert engels <re...@ix.netcom.com>.
It is not just database by the way, any journaling file system would
be pointless...
On Nov 19, 2008, at 12:55 PM, robert engels wrote:
> The "utility" referenced no longer exists... and its no wonder.
>
> If is most likely that the tester did not have the drives
> configured properly.
>
> In almost all cases, if the drive did this, you could not run a
> database system with any resiliency.
>
> They would also have problems with shutdown - although as the
> comments indicate Windows added a "delay" to try and fix this.
>
> There is a chance that these drives just had bugs that caused the
> cache to not be written properly, but that is probably very rare.
>
> More likely, they were testing higher end drives that have this as
> an option. If your system is fault tolerant, with back power
> supplies, etc. then having the drive do lazy writing, even in
> response to sync, can drastically improve the performance - but you
> need to make sure the rest of the system is configured for this.
>
> I guess if you buy completely crappy hardware, or a bad system
> configuration, you get what you pay for.
>
>
> On Nov 19, 2008, at 11:44 AM, Jason Rutherglen wrote:
>
>> http://www.h2database.com/html/advanced.html#durability_problems
>> http://hardware.slashdot.org/article.pl?sid=05/05/13/0529252
>>
>> On Wed, Nov 19, 2008 at 9:23 AM, Mark Miller
>> <ma...@gmail.com> wrote:
>> /A look at the slashdot article seems to indicate the OS may have
>> been at fault in his case (an update he posted in response to some
>> slashdot flames). Thats interesting. Wasnt there the last time
>> this topic came up. Can't remember what java db had the other
>> info...but looking...
>>
>> /
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: java-dev-unsubscribe@lucene.apache.org
>> For additional commands, e-mail: java-dev-help@lucene.apache.org
>>
>>
>
Re: Option to fsync files
Posted by robert engels <re...@ix.netcom.com>.
The "utility" referenced no longer exists... and its no wonder.
If is most likely that the tester did not have the drives configured
properly.
In almost all cases, if the drive did this, you could not run a
database system with any resiliency.
They would also have problems with shutdown - although as the
comments indicate Windows added a "delay" to try and fix this.
There is a chance that these drives just had bugs that caused the
cache to not be written properly, but that is probably very rare.
More likely, they were testing higher end drives that have this as an
option. If your system is fault tolerant, with back power supplies,
etc. then having the drive do lazy writing, even in response to sync,
can drastically improve the performance - but you need to make sure
the rest of the system is configured for this.
I guess if you buy completely crappy hardware, or a bad system
configuration, you get what you pay for.
On Nov 19, 2008, at 11:44 AM, Jason Rutherglen wrote:
> http://www.h2database.com/html/advanced.html#durability_problems
> http://hardware.slashdot.org/article.pl?sid=05/05/13/0529252
>
> On Wed, Nov 19, 2008 at 9:23 AM, Mark Miller
> <ma...@gmail.com> wrote:
> /A look at the slashdot article seems to indicate the OS may have
> been at fault in his case (an update he posted in response to some
> slashdot flames). Thats interesting. Wasnt there the last time this
> topic came up. Can't remember what java db had the other info...but
> looking...
>
> /
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: java-dev-unsubscribe@lucene.apache.org
> For additional commands, e-mail: java-dev-help@lucene.apache.org
>
>
Re: Option to fsync files
Posted by Mark Miller <ma...@gmail.com>.
Ummm...well durability isnt really corruption is it. I guess I'm convert
there.
Mark Miller wrote:
> Thanks Jason...had just found h2 again myself - that guy seems to say
> even with write caching disabled on the OS, its still a problem
> (though probably less of a problem than without sync?
>
> Jason Rutherglen wrote:
>> http://www.h2database.com/html/advanced.html#durability_problems
>> http://hardware.slashdot.org/article.pl?sid=05/05/13/0529252
>>
>> On Wed, Nov 19, 2008 at 9:23 AM, Mark Miller <markrmiller@gmail.com
>> <ma...@gmail.com>> wrote:
>>
>> /A look at the slashdot article seems to indicate the OS may have
>> been at fault in his case (an update he posted in response to some
>> slashdot flames). Thats interesting. Wasnt there the last time
>> this topic came up. Can't remember what java db had the other
>> info...but looking...
>>
>> /
>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: java-dev-unsubscribe@lucene.apache.org
>> <ma...@lucene.apache.org>
>> For additional commands, e-mail: java-dev-help@lucene.apache.org
>> <ma...@lucene.apache.org>
>>
>>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-dev-help@lucene.apache.org
Re: Option to fsync files
Posted by Mark Miller <ma...@gmail.com>.
Thanks Jason...had just found h2 again myself - that guy seems to say
even with write caching disabled on the OS, its still a problem (though
probably less of a problem than without sync?
Jason Rutherglen wrote:
> http://www.h2database.com/html/advanced.html#durability_problems
> http://hardware.slashdot.org/article.pl?sid=05/05/13/0529252
>
> On Wed, Nov 19, 2008 at 9:23 AM, Mark Miller <markrmiller@gmail.com
> <ma...@gmail.com>> wrote:
>
> /A look at the slashdot article seems to indicate the OS may have
> been at fault in his case (an update he posted in response to some
> slashdot flames). Thats interesting. Wasnt there the last time
> this topic came up. Can't remember what java db had the other
> info...but looking...
>
> /
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: java-dev-unsubscribe@lucene.apache.org
> <ma...@lucene.apache.org>
> For additional commands, e-mail: java-dev-help@lucene.apache.org
> <ma...@lucene.apache.org>
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-dev-help@lucene.apache.org