[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: ext3 and data=journal bug


On Thu, 2003-05-01 at 15:57, Mark A Basil wrote:
> Actually, now that I think about it, the major constraint is the disk space 
> needed for a >200MB journal.  Most servers do have the space, but this will 
> be a network wide migration to ext3, and some servers are somewhat limited.
> I'm now beginning to question the journal mode, which is disappointing due to 
> the speed advantage it provides.  Would it be extremely unwise to run a 
> journal of 50MB in data=journal?  

Not at all.  

> Would it help to flush the buffer every 10 
> seconds instead of the default of 30?  

Actually, the default for the journal is 5 seconds, but in this case it
doesn't matter --- a journal commit is required every single time you
encounter a synchronous IO, so the sorts of situations where
data=journal helps are situations where you're getting a forced commit
several times a second anyway.  Changing the default commit timeout
really doesn't affect those situations at all.


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]