[DRBD-user] Multi TB filesystems

Leroy van Logchem Leroy.vanLogchem at wldelft.nl
Wed Oct 12 16:58:58 CEST 2005

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


> Is drbd fine with that amount of diskspace? What caveats will I face? 
> Your experience?

In my experience one can use drbd using multi TB filesystems. No
flamebaits involved
but the filesystem type can make it stable or unstable. Also you really
want to keep the
meta-data seperate to increase your options when disaster strikes. This
will keep the
on-disk structure identical to a none drbd and hence you can even mount
it without drbd.
Again, only as last resort *if* you're in big troubles. Since using drbd
for more then six
months on multiple multi TB servers I never lost data. Among the servers
are different
storage backends (scsi connected sata cabinets) and directly connected
sata disks using
Areca pci-express controllers. All this is my personal view but here we go:

+++ Filesystem must be rock stable: ext3 with both meta and block journal
++ Lower the nr. of inodes, in context of ext3 (mkfs.ext3 -i 102400 for
example )
+Use the meta-disk seperate partition option ( meta-disk /dev/sdc2[0]; )
+Use 9000 bytes of MTU size using high quality nics ( Intel PRO ).

Over the last years, actualy using, large filesystems on xfs/reiser
always gave
troubles [ filesystems totallty wracked, inaccesible files etc etc. ].

Filesystems I trust are: vxfs on Solaris, ufs2 on FreeBSD and ext3 on Linux.

We got five clusters running with filesystems like:
/dev/drbd0 1.4T  920G  436G  68% /part1
/dev/drbd1 1.4T  963G  393G  72% /part2
/dev/drbd2 1.4T  1.3T     93G  94% /part3

--Leroy




More information about the drbd-user mailing list