[NCLUG] Re: ext3 vs. ReiserFS

John L. Bass jbass at dmsd.com
Wed Oct 24 16:49:56 MDT 2001


	I know for a fact that availability of dump/restore would make or break
	the adoption of a specific filesystem where I work... :)

I've been using dump w/amanda+EXB-440 on several ext3 filesystems under the RH7.1.94 beta.
The dump see's the journal file and appears to do the right thing on lightly loaded
filesystem. I haven't needed (wanted to spend the time yet) to stress test it. I may
soon just to be safe.

I have been using two large software Raid 5 arrays - FC 6x18GB and 12x9GB 10K Drives.
There are serious problems with the raid software during power sags where the supply
might drop a bit and reset a few of the active drives. The Linux raid software makes
a huge mistake with this, and varies the drives off line and attempts to briefly
run with fewer than enough drives. The result is the serial/transaction numbers in the
array get out of wack and the array is useless until manually recovered. I could probably
hide this with a faster UPS and more lightly load the drive array power - but it's
a problem in any case.

Will be upgrading all 4 systems tonight/tomarrow to RH7.2 to get off the beta which
has been pretty stable.

I'm seriously thinking about switching to ReiserFS - but all these filesystems are
far less robust than they try to imply. They only journal meta data at best, which
really leaves the filesystem wide open for serious data corruption that might take
months to notice (past normal full dump holding cycles for many shops).

John



More information about the NCLUG mailing list