quotas on UFS file systems
Liam Hoekenga
addition to user based quotas) in Solaris UFS based filesystems? I've
done some looking around, and I don't think it's possible, but my
manager is pushing hard for it.
Are there other filesystems (prolly commericial) that offer this
functionality?
Liam
Mark Montague
As far as I know, no local filesystem on Solaris supports this
functionality (I even looked on our Solaris 8 development machines).
If you are going to stick to Solaris, I'd recommend using AFS so that
you can set the quota on the volume and then only give the particular
group you're interested in write ACLs to the volume. A large AFS cache
can improve performance and minimize (to the extent possible)
performance penalties, and if you're really interested in speed, then
put a lot of RAM in the computer and configure the cache to be a
RAM-based cache rather than a disk-based cache (how to do this was
discussed on the info-afs mailing list last month).
Linux, however appears to allow quotas for groups:
NAME
setquota - set disk quotas
SYNOPSIS
/usr/sbin/setquota [ -n ] [ -u | -g ] name filesystem-name
block-softlimit block-hardlimit inode-softlimit inode-
hardlimit
/usr/sbin/setquota [ -n ] [ -u | -g ] [ -p protoname ]
name filesystem-name
DESCRIPTION
setquota is a command line quota editor. The filesystem,
user/group name and new quotas for this filesystem can be
specified on the command line.
-n Edit also non-local quota use rpc.rquotad on remote
server to set quota.
-u set user quotas for named user. This is the default
-g set qroup quotas for named group
[...]
NAME
quota - display disk usage and limits
SYNOPSIS
quota [ -guv | q ]
quota [ -uv | q ] user
quota [ -gv | q ] group
DESCRIPTION
Quota displays users' disk usage and limits. By default
only the user quotas are printed.
-g Print group quotas for the group of which the user
is a member. The optional
[...]
This is from a Red Hat 6.2 system.
Mark Montague
LS&A Information Technology
mark...@umich.edu
Charles Clark
|Does anyone have suggestions for implementing group based quotas (in
|addition to user based quotas) in Solaris UFS based filesystems? I've
Depending upon how the filesystem tree is layed out, you could maybe
manage it with disk partitions. If the various groups need to all throw
their files into the same pile then no. If each group can have its own
part of the tree carved out, then you can fake quotas with disk
partitions (hard to resize though!) or vxvm (veritas volume manager)
volumes. disksuite also works, sort of, mangling the benefits of vxvm
volumes (to a certain degree) with the problems of dealing with disk
partitions.
|done some looking around, and I don't think it's possible, but my
|manager is pushing hard for it.
ufs has bsd style (ie per user) quotas.
|Are there other filesystems (prolly commericial) that offer this
|functionality?
The only other filesystems I've used (on a Sun) are afs, nfs, and vxfs
(veritas file system). vxfs is a local disk filesystem (same as ufs)
and also only has bsd style quota support. nfs has quota support, but I
don't know any nfs servers that let you give quotas based on unix
groups. netapp quota trees are a quota on a directory and all its
children (functionally the same mounting a resizable filesystem at that
directory for diskspace usage limiting purposes).
I don't know if afs would give you (pts) group quotas. I've been away
from it too long. This obviously doesn't help you with your local disk,
which may or may not matter to you.
--
cmc ACGWB #1 PAW #4 DoD #1325
"Drove back to town this morning with working on my mind
I thought of maybe quitting, thought of leaving it behind
went back to bed this morning, and as I'm pulling down the blind
the sky was dull and hypothetical and falling one cloud at a time."
-- The Tragically Hip
Charles Clark
|
|Linux, however appears to allow quotas for groups:
cool; you could implement this by nfs-serving from linux then. users on
the solaris nfs client would have no way to check their group quota
usage on the server via native solaris unix tools though. Until they
hit the limit, that is :)