cost of tdb transactions on ext4 with barriers

tridge at tridge at
Fri Feb 12 04:25:18 MST 2010

 > I would assume that fsync() is all I need in order to know confidently
 > that my data and metadata will always survive a crash.

I used to think that until Rusty pointed out the error of my ways
with his "pull the plug" testing with and without ext3 barriers :)

More information about the samba-technical mailing list