On 11/12/97 16:13:05 you wrote:
I have answered this question many times, and I am sick of it.
Hey, you don't have to convince *me*. I didn't mean it as a personal attack. But I do know it has been a source of confusion. As you see yourself, different people here had different expectations. So perhaps it's not quite as obvious as it seems?
Go onto your own SunOS or Solaris system and try it yourself, soft quotas have been around for a very long time and so far, I have not heard a single technical reason why it should not work.
My investigations into this tell me they work much like Guy Harris says. Do you agree with his assessment? If your version of soft quotas differs, in what way does it differ? Does your SunOS system report you as over soft quota over NFS as soon as it happens?
I'm only trying to facilitate the questions and answers here.
FreeBSD does it, shoddy Linux does it, hell, I can't name a unix system off the top of my head that doesn't.
Is it so much to ask?
What would a Netapp implementation do, other than report a "time left" or somesuch in the field reported via the quota command?
Bruce
On Wed, Nov 12, 1997 at 04:40:46PM -0600, sirbruce@ix.netcom.com had written:
On 11/12/97 16:13:05 you wrote:
I have answered this question many times, and I am sick of it.
Hey, you don't have to convince *me*. I didn't mean it as a personal attack. But I do know it has been a source of confusion. As you see yourself, different people here had different expectations. So perhaps it's not quite as obvious as it seems?
Sorry for jumping down your throat so quickly, that was wrong. My frustration with this is with Network Appliance, not with you.
But, it is obvious, if the engineering folx at NetApp actually decided to try it out.
Go onto your own SunOS or Solaris system and try it yourself, soft quotas have been around for a very long time and so far, I have not heard a single technical reason why it should not work.
My investigations into this tell me they work much like Guy Harris says. Do you agree with his assessment? If your version of soft quotas differs, in what way does it differ? Does your SunOS system report you as over soft quota over NFS as soon as it happens?
I do not agree with his assessment.
Yes, it does report it.
I'm only trying to facilitate the questions and answers here.
I understand.
Again, I am not alone. I don't know why the others on this list are being quiet about it, but alas they are. SPEAK UP LURKERS!
FreeBSD does it, shoddy Linux does it, hell, I can't name a unix system off the top of my head that doesn't.
Is it so much to ask?
What would a Netapp implementation do, other than report a "time left" or somesuch in the field reported via the quota command?
Nothing except return warning messages via the quota daemon or something.
Like I said, SunOS, Solaris, Linux, FreeBSD, etc, all work fine.
On Wed, 12 Nov 1997 16:56:28 CST, Mike Horwath wrote:
Again, I am not alone. I don't know why the others on this list are being quiet about it, but alas they are. SPEAK UP LURKERS!
*lurk* *lurk* *slither* *slither*
--- Brett Rabe Email : brett@uswest.net Systems Administrator - !nteract Services Phone : 612.664.3078 600 Stinson Blvd. Pager : 612.613.2549 Minneapolis, MN USA 55413 Fax : 612.664.4770
If you aren't the lead dog, the view is always the same.
I'm not an expert in this area, but there is supposedly a definitive technical limitation with NFS limiting support of soft quotas. The architects of NFS did not implement a return call to indicate "I wrote your file, but..."
The only return codes from a NFS server indicate success or failure of the request. You would have to lobby with the architects of both NFS client and server software to ask them to add this for you - Netapp is only on one side of this coin.
The previously posted idea of periodic "quota reports" and sending email or other nastygrams to "soft limit" offenders is not a bad idea, although not as nice as what we all really want.
Realistically, messages at login are almost always ignored. Most of our users login once per month or so and just lock their screens nightly. You really need more frequent notification than once at login, and that is what soft quotas give you, so be carefull for what you ask...
We're playing in a standards-based arena, and are on the field with zillions of other teams. It's good to ask for improvements, but change is likely to be slow.
I can see one DISadvantage of soft quotas. When a user passes the soft quota he sees at most one diagnostic message at that moment and may see nothing at all. If login is configured, the user gets his disk quota status at login, but many users ignore login messages. If the hard limit is much larger than the soft limit, the user may consume quite a bit more disk during his one week grace period (but still stay below the hard limit). When the grace period runs out, the system does not let him have a single disk block until he deletes enough files to get back under his SOFT limit. Our help desk staff (who do not have superuser and cannot increase quotas) occasionally have to deal with very confused and frustrated users who fall into this trap, even though we send out automatic email warnings to users who are over their soft limit.
"I couldn't save this file I was editing, so I freed up more than enough disk space and it still won't let me save. What's that? You're kidding. I have to delete 50 more Mbytes before I can save this file?"
We have some crafty users who routinely stay above their soft limit. When their grace period runs out, they simply copy all their stuff to /tmp, clean out their home directory to reset the clock, and copy everything back. They're good for another week.
Snapshots would make this trick even easier -- no copying to /tmp needed. Just clean out your directory and copy everything back from the latest snapshot.
Don't get me wrong -- I'm sure that if NetApp were to support soft quotas, we would use them. But we can live without the soft quotas. We'll send out automatic email warnings to users close to their limit.
Steve Losen scl@virginia.edu phone: 804-924-0640
University of Virginia ITC Unix Support
Steve Losen penned:
I can see one DISadvantage of soft quotas. [Every concievable way to get around quotas deleted.]
Every issue you raise is completely correct. Since I administer several community access sites, I have the luxury of being a disk fascist when the two week grace period expires. I get to delete files in descending order of size until the account is below its quota. I can't delete saved email though, so a report is generated for the helpdesk of whose disk use is frozen due to excessive saved email. This works beautifully for sites with soft quotas, but does not work on our newest site which uses a NetApp for home directories. I'm living with it, but it definitely is one thing I would like to see (along with ssh access) in a future OS.
Regards,
David K. Drum david@more.net
Nothing except return warning messages via the quota daemon or something.
The quota daemon, if by that you mean "rquotad", reports things only when asked. It does not spontaneously send messages to NFS clients informing them that user XXX has gone over quota.
Have you seen traffic on the wire indicating that it's asked any time other than when the "quota" command is run? (It gets run, on many systems, at login time.)
It's straightforward for us to have our quota daemon report soft quota information. Modulo changing the quota file format (as we're out of space), it's pretty much straightforward for us to keep soft quota information in the quota file. Keeping it in memory could chew up more memory, but if we're keeping the entire quota database in memory, I'd consider that a deficiency that should be fixed (although it might take some effort to fix it).
It is *not* straightforward to have the user who goes over quota notified immediately if they go over quota on a file system ounted from a remote NFS server, which is why SunOS 4.1.3 didn't do it (and why 5.x doesn't do it either).
-
Brett Rabe -
David Drum -
Fritz Feltner -
guy＠netapp.com -
Mike Horwath -
sirbruce＠ix.netcom.com -
Steve Losen