[BBLISA] Diskspace monitoring suggestions.

Dean Anderson dean at av8.com
Sun Sep 12 17:45:59 EDT 2004


It depends on what you mean by 'monitoring'.  Do you want to know how much
apace is used by a given tree (say user quotas), or do you want to know
when you'll run out of space?

Du is probably not the right tool, and probably the definitely wrong tool
to use over NFS.  But it depends on what you are trying to find out.  Du
recursively walks a tree. But df tells you directly how much space is used
on a partition.  Often you don't really care about the tree so much as you
care about when the partition is full and there is no more space
available.  Other times, smart partitioning allows you convert the former
into the latter.

Depending on what you are trying to accomplish, quotas may be a friend. Or
lvm/vxvm/etc might allow you to create a very large number of partitions.  
One partition per user can be handy to manage space (eg afs sites usually
do this), but hard on the size of the mount table (this isn't a problem
for afs, but might be for NFS or lvm mounts on the fileserver).

For collecting and viewing data and trends, I'd suggest that you look at
MRTG and Naggios.


		--Dean

On Sun, 12 Sep 2004, Michael C Tiernan wrote:

> I'm wondering if anyone has any suggestions on how to do this. I can't belive 
> I'm the first or only person to run into this situation.
> 
> I've got multiple big disk arrays that are available via NFS to multiple 
> systems.  On these arrays are "areas" where multiple users store lots of 
> data.
> 
> I need to monitor this on a regular/periodic basis. I've also got to generate 
> a history of changes of these different parts of the directory trees.
> 
> Before suggestions of simple 'du' commands are offered, understand that a 'du' 
> on one of these trees currently takes 13 hours. Not real efficient nor does 
> it qualify as "snapshots". :)
> 
> Does anyone have any suggestions for approaches or tools that can make this 
> work better?  Any opinions?
> 
> As you can imagine, I'm limited in what I can and can't change about the 
> structures I'm working in.
> 
> Thanks in advance for everyone's time.
> 
> _______________________________________________
> bblisa mailing list
> bblisa at bblisa.org
> http://www.bblisa.org/mailman/listinfo/bblisa
> 
> 




More information about the bblisa mailing list