One question that keeps coming up in my discussions about ZFS is the lack of user quotas. Typically this comes from people who have many tens of thousands (30,000 - 100,000) of users where they feel that having a file system per user will not be manageable. I would agree that today that is the case however I personally don''t see this as an insurmountable problem. However the questions keep being asked: Could user quotas be introduced on ZFS? Are there any plans to do this? This message posted from opensolaris.org
Chris Gerhard wrote:> One question that keeps coming up in my discussions about ZFS is the lack of user quotas. > > Typically this comes from people who have many tens of thousands (30,000 - 100,000) of users where they feel that having a file system per user will not be manageable. I would agree that today that is the case however I personally don''t see this as an insurmountable problem. However the questions keep being asked:Given how quotas are managed on UFS are these people really having 30,000 entries in the quota file ? When all we had was the UFS qutoa system I kept hearing that people didn''t like it and what they really wanted was group level quotas. What ZFS gives you is a nice easy way to do group level quotas, you give a group a filesystem and you have group quotas. I see the biggest problem with qutoas at the user level being how to effectively manage the data. Particularly if more than a single file system is in use and most especially if there is more than a single host involved. From the people that keep asking for this, how would they like to see this managed in an idea world ? -- Darren J Moffat
Chris Gerhard writes: > One question that keeps coming up in my discussions about ZFS is the lack of user quotas. > > Typically this comes from people who have many tens of thousands > (30,000 - 100,000) of users where they feel that having a file system > per user will not be manageable. I would agree that today that is the > case however I personally don''t see this as an insurmountable > problem. However the questions keep being asked: > > Could user quotas be introduced on ZFS? > > Are there any plans to do this? > I hope not. I''d rather see the resources directed at fixing the issues associated with managing 100000 FS (RFE #?) than producing a temporary wart (that will need to be managed itself, with bugs, RFEs, etc). It''s understandable that people would like to fit ZFS into their mental model but I don''t think we should give up that easily particularly if we think the ZFS model is fundamentally better. If we can fix the problems and end up in a better situation for everybody, I think we should focus on that. -r > > This message posted from opensolaris.org > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Roch - PAE wrote:> Chris Gerhard writes: > > > One question that keeps coming up in my discussions about ZFS is the lack of user quotas. > > > > Typically this comes from people who have many tens of thousands > > (30,000 - 100,000) of users where they feel that having a file system > > per user will not be manageable. I would agree that today that is the > > case however I personally don''t see this as an insurmountable > > problem. However the questions keep being asked: > > > > Could user quotas be introduced on ZFS? > > > > Are there any plans to do this? > > > > I hope not. I''d rather see the resources directed at fixing > the issues associated with managing 100000 FS (RFE #?) > than producing a temporary wart (that will need to be > managed itself, with bugs, RFEs, etc). > > It''s understandable that people would like to fit ZFS into > their mental model but I don''t think we should give up that > easily particularly if we think the ZFS model is > fundamentally better. > > If we can fix the problems and end up in a better situation > for everybody, I think we should focus on that. > > -r > > >Nobody likes managing quotas -- it is cumbersome, ungainly and error prone. However, it *is* an operational requirement at many facilities and a means of exercising control over the namespace. Why would you expect that large data-rich institutions invested in this technology would just jump ship and embrace your "mental model" when your solution is so constrained? Tens of thousands of filesystems exported via NFS? Earlier I posted a scenario where this is deficient. Managing thousands of networked filesystems sensibly without user quotas is not a new idea -- AFS or Panasas both do this by caching files on a local filesystem without requiring hundreds of client mounts ( unfortunately, this would require you to provide a hook into the client kernel ). -- Robert Petkus Brookhaven National Laboratory Physics Dept. - Bldg. 510A Upton, New York 11973 http://www.bnl.gov/RHIC http://www.acf.bnl.gov