Journaling filesystems

[dsr+inn at mail.lns.cornell.edu]

"Jeffrey M. Vinocur" <jeff at litech.org> writes:
> We're now running with journaling filesystems.

Which one?  Many "journaling" filesystems (including apparently most
of the ones for Linux) only journal metadata, not data (journaling
data can be significantly slower).

> As a result, it seems that no matter how much msync'ing we do,
> nothing truly gets written except on a periodic basis (say, every 30
> seconds).  I don't even know if closing the file guarantees that it
> will be written.

Closing the file should[1] force a synchronous update.  So should[1]
an msync with the MS_SYNC flag.

> Any ideas on what to do about this?  It seems like a major step backwards 
> for us, even if it does guarantee the integrity of the filesystem.

Use a filesystem that journals data, convert some of the MS_ASYNCs to
MS_SYNC (may require substantial meditation on the code to identify
the right subset of msync()s), or twiddle some of the parameters to
update(8) (may have a different name depending on your OS) to see if
you can reduce the data flush time to something less than 30 seconds.
(Listed in decreasing order of useful data consistency properties.)

[1] no gurantee it actually does.  However, IMO, any decent filesystem
that doesn't make close synchronous should make that behavior a mount
option.
-- 
"about 15 percent of the people are screwballs, lightweights and boobs
   and you would not want those people unrepresented in Congress."
	    -Alan Simpson, former US Senator (R, Wyoming)