default ext3 journal mode - time to poke at this again?

default ext3 journal mode - time to poke at this again?

[Ric Wheeler]

At the linux symposium this week, some of us were wondering when it 
would be safe to try and default ext3's default journal mode back to 
sanity given that all (?) distros immediately reverted the default to 
data writeback mode.

Failing that, can we at least clean up the config option description to 
properly give the normal users a non-bogus, technical description of the 
trade off ?

Thanks!

Ric

Re: default ext3 journal mode - time to poke at this again?

[Jan Kara]

  Hi,

> At the linux symposium this week, some of us were wondering when it 
> would be safe to try and default ext3's default journal mode back to 
> sanity given that all (?) distros immediately reverted the default to 
> data writeback mode.
  Well, at least SUSE still uses data=ordered,barrier as a default (and
I personally think this is the right thing to do). Hmm, maybe you've
meant this and just wrote it otherwise given your paragraph below ;)

> Failing that, can we at least clean up the config option description to 
> properly give the normal users a non-bogus, technical description of the 
> trade off ?
  Yes, I'm all for it. Will you try to write up something or should I?

									Honza
-- 
Jan Kara <jack@suse.cz>
SuSE CR Labs

Re: default ext3 journal mode - time to poke at this again?

[Ric Wheeler]

On 07/20/2009 02:57 PM, Jan Kara wrote:
>    Hi,
>
>    
>> At the linux symposium this week, some of us were wondering when it
>> would be safe to try and default ext3's default journal mode back to
>> sanity given that all (?) distros immediately reverted the default to
>> data writeback mode.
>>      
>    Well, at least SUSE still uses data=ordered,barrier as a default (and
> I personally think this is the right thing to do). Hmm, maybe you've
> meant this and just wrote it otherwise given your paragraph below ;)
>    

Definitely what I meant to post - Fedora and RHEL both use ordered mode 
as a default as well. Not many fs developers are fans of data writeback 
mode as far as I can tell...

>    
>> Failing that, can we at least clean up the config option description to
>> properly give the normal users a non-bogus, technical description of the
>> trade off ?
>>      
>    Yes, I'm all for it. Will you try to write up something or should I?
>
> 									Honza
>    

I posted something (from rwheeler@redhat.com) earlier today and we had a 
discussion about this on the ext4 call. I think that Ted will try to 
update the text that I proposed over the next couple of days.

I think that Chris will be reposting his data guarded mode patches soon 
as well, so we might be able to move forward quickly :-)

ric