[Linux-cluster] GFS2 file system maintenance question.
jduston at ll.mit.edu
Tue Mar 15 20:55:27 UTC 2011
These certainly are concerns, although straying a little off-topic. I
am risk-averse and try to avoid or mitigate 'gotcha' issues. (Hence why
I'm thinking about maintenance issues now). Unfortunately, having less
data is not an option, so I need to try to implement the best solution
available to handle that data.
I'm not sure backing up, say 5 x 100TB filesystems would be much
difference from backing up the same data on a single 1 x 500TB filesystem.
Restoring a single 100TB filesystem would definitely be easier than a
single 500TB filesystem, but the trade-off is that the data is split up
across 5 filesystems.
Tape backup has definite drawbacks when you start dealing with large
data sets. Fortunately our data will not be changing often. We are
backing up to external hard drives using trayless chassis, basically
using 2TB HDDs as jumbo floppy drives. YMMV.
We are presently running a 70TB XSan2/StorNext SAN. It has been
rock-steady since created, about half a year now (XSan1, not so much).
I hope creating a 100TB filesystem from one designed to scale to 8EB
should really not be too much of a test for GFS2. That is only 1/80th
its design capacity. I do not think the developers at Red Hat are any
less capable than those at Quantum. (although it will certainly suck to
uncover an edge case or bug triggered by >100TB filesystem!).
Given the choice, I certainly would not be pushing the boundaries of
what's officially supported.
However, it does seem Red Hat has built a Monster Truck. Since I have
cars that need crushing, lets see if it can crush some cars before just
parking it in the drive.
p.s. I'll start building more driveways if it can't, but lets at least
try it first...
On 03/15/2011 02:26 PM, Alan Brown wrote:
> Jack Duston wrote:
>> Thanks Yue, but your information would seem dated if this site is correct:
>> Even if 100TB is what's officially supported in RHEL6, it doesn't mean
>> that larger file systems won't work.
> Anyone considering such large filesystems should consider the following
> 1: How long is it going to take to back it up.?
> 2: How long will it take to restore??
> Even LTO5 takes the best part of 12 hours to restore 1Tb...
> Linux-cluster mailing list
> Linux-cluster at redhat.com
More information about the Linux-cluster