toasters

toasters@lists.teaparty.net

13515 discussions

RE: Max number of drives.
by Thompson, Arnold 07 Oct '99

07 Oct '99

I recall from mixing 9s and 18s at a customer site that you *can* add the 18s to an existing volume of 9s and still get 18G worth (minus filesystem overhead, snaps, etc.) from each 18G drive. The first 18G drive becomes parity and the 9G parity becomes a data drive (so you get 9G additional capacity from the first 18G drive), but after that you should get 18 and not 9. Historically, you've been able to mix and match with the filer doing the right thing. It is true that if a 9G drive fails and there is no 9G spare, it will rebuild on an 18G drive but you will only get 9G on that rebuilt drive. So be sure to have a spare of each size available. If anybody has tested this with different results I could be completely nuts, so please educate us! It is important to make the 9G shelves the lowest order shelves for FCAL loop initialization purposes, so make the 9G shelves numbers 0 and 1. Regards, Arnie > -----Original Message----- > From: Mohler, Jeff [mailto:jeff.mohler@netapp.com] > Sent: Thursday, October 07, 1999 11:38 AM > To: 'Jason Middlebrooks'; toasters(a)mathworks.com > Subject: RE: Max number of drives. > > > Sure, no problem. > > But when you add them to a volume (if you are not creating a > new one) make sure > you add them to the volume made up of 18G drives. (That is, > if the 18G drives > are a separate volume from the 9G ones). > > If you mix/max drive sizes in a volume, the larger drives > will appear only to be > the size of the smallest drive in the array. > > -----Original Message----- > From: Jason Middlebrooks [mailto:Jason_Middlebrooks@datalink.com] > Sent: Thursday, October 07, 1999 12:01 PM > To: toasters(a)mathworks.com > Subject: Max number of drives. > > > Question. > > I have 14 9Gig drives and 14 18Gig drives. So a total of 28 > drives. Can I > add 7 more 18Gig drives on the same loop? > > The filer is a F760 > > Thanks in advance > Jason Middlebrooks >

1 0

RE: Max number of drives.
by Mohler, Jeff 07 Oct '99

07 Oct '99

Sure, no problem. But when you add them to a volume (if you are not creating a new one) make sure you add them to the volume made up of 18G drives. (That is, if the 18G drives are a separate volume from the 9G ones). If you mix/max drive sizes in a volume, the larger drives will appear only to be the size of the smallest drive in the array. -----Original Message----- From: Jason Middlebrooks [mailto:Jason_Middlebrooks@datalink.com] Sent: Thursday, October 07, 1999 12:01 PM To: toasters(a)mathworks.com Subject: Max number of drives. Question. I have 14 9Gig drives and 14 18Gig drives. So a total of 28 drives. Can I add 7 more 18Gig drives on the same loop? The filer is a F760 Thanks in advance Jason Middlebrooks

2 1

Max number of drives.
by Jason Middlebrooks 07 Oct '99

07 Oct '99

Question. I have 14 9Gig drives and 14 18Gig drives. So a total of 28 drives. Can I add 7 more 18Gig drives on the same loop? The filer is a F760 Thanks in advance Jason Middlebrooks

1 0

RE: Max number of drives.
by Alvarado, Michael 07 Oct '99

07 Oct '99

You can add the additional 7 disks since that will still be less than the maximum storage capacity currently supported on an F740. The maximum capacity of an F740 is 928 GB. -----Original Message----- From: Jason Middlebrooks [mailto:Jason_Middlebrooks@datalink.com] Sent: Thursday, October 07, 1999 12:01 PM To: toasters(a)mathworks.com Subject: Max number of drives. Question. I have 14 9Gig drives and 14 18Gig drives. So a total of 28 drives. Can I add 7 more 18Gig drives on the same loop? The filer is a F760 Thanks in advance Jason Middlebrooks

1 0

More Backup Questions
by Jay Orr 07 Oct '99

07 Oct '99

Hi. Still curious about this all. Thanks for all the info people have sent me. Here's some more questions. I'm working from a system where we have a jukebox hanging off a sun box backing up an F330 (well, there's more then this, but I'm keeping it simple). 1) Does the current (5.3.2r1) version support NDMP v3? a) If not is v3 support planned and for when? 2) What (if any) advantage is there using NDMP vs. Dump (dump.FAServer)? 3) Which is better/faster for individual file restores (i.e. one or two files only)? 4) Can you extract a file out of a NDMP dump onto a different file system (i.e. onto a sun box)? 5) Are both snapshot backups (NDMP & dump.FAServer)? Important for ClearCase users.... 6) Any problems spanning NDMP dumps over multiple tapes? Concerns that bring me to these questions : 1) Budtool (now Legato) not wanting to support dump.FAServer types. vs 2) End of Life on the F330's. ----------- Fujitsu - Nexion, St. Louis, MO Jay Orr (314) 579-6517

2 1

Re: More Backup Questions
by Darrell Fuhriman 07 Oct '99

07 Oct '99

Jay Orr <orrjl(a)stl.nexen.com> writes: all of my answers assume you're backing up to local tape. > 1) Does the current (5.3.2r1) version support NDMP v3? No. the standard has not been finalized and to my knowledge no one has implemented it. > a) If not is v3 support planned and for when? I'm trying to find the answer to this myself. > 2) What (if any) advantage is there using NDMP vs. Dump (dump.FAServer)? well, NDMP will work with your backup software better. Keep in mind, NDMP is nothing but a dump which spits back some meta information to the backup server. > 3) Which is better/faster for individual file restores > (i.e. one or two files only)? they're equally bad for the reason given above. > 4) Can you extract a file out of a NDMP dump onto a different file system > (i.e. onto a sun box)? you should be able to, though frankly I've never tried it. > 5) Are both snapshot backups (NDMP & dump.FAServer)? Important for > ClearCase users.... yes. > 6) Any problems spanning NDMP dumps over multiple tapes? no. Darrell

1 0

HTTP on a filer
by mitch＠netline.com 06 Oct '99

06 Oct '99

Greetings. I hope there are some folks out here that have experience using a filer as an HTTP server. First some specifics: Filer: F740, 512MB RAM, 32MB NVRAM, FC disks, 100Mbit switched Enet conn. Version: NetApp Release 5.3.2R1: Thu Aug 12 19:16:27 PDT 1999 We purchased the HTTP protocol license to offload images to the filer directly rather than have an intermediate host serve them up by grabbing them off that filer. After some basic testing we put this into the start of production and all was well, though first connections by a browser seemed to lag a bit, subsequent connections were fine. All was well for about 2 weeks when things came to a crawl and finally to a stop. Several observations 1. httpstat reported 512 peak and 512 open connections. 2. despite doing an "options httpd.enable off" and leaving it down for a while, open conn. never dropped and the http server was non-responsive. I even tweak the timeout down to 30, which is the lowest it can go. 3. NFS service was seemingly unaffected 4. netstat reported *many* (512?) connections in either ESTABLISHED or CLOSE_WAIT and those connections didn't seem to ever go away. 5. The log file was at it's maximum. I moved it to another filename and did an httpd.enable on then off, but another log file was never written to. How do you rotate the logs on a filer for HTTP? A possible kink in the matter is that I am using the VH interface and have two IP's aliased to it, the filer's address is aaa.bbb.ccc.xxx and then it has: ifconfig vh alias aaa.bbb.ccc.yyy ifconfig vh alias aaa.bbb.ccc.zzz I then use the httpd.hostprefixes to map each VH to the appropriate area: /path/YYY host-y.domain.com aaa.bbb.ccc.yyy /path/ZZZ host-z.domain.com aaa.bbb.ccc.zzz /path/TESTy test-y.domain.com /path/TESTz test-z.domain.com and this all works fine as far as things being pointed to the right places. All of this leaves me with some questions: 1. How do you turn off httpd on a filer? "options httpd.enable off" does not seem to do the trick. After doing that, if I telnet to port 80 I do not get a connection refused, it just sits there. At least when it was in this "hung" state. 2. How do you rotate log files. Actually, for the images, I don't want the log file at all, but the manual doesn't say how to make it go to the filer equivalent of /dev/null. 3. Can the number of connections be raised from 512? Can the timeout be made shorter than 30 seconds? Can the filer close idle/lost connections faster? 4. Is this a known problem with a fix? NOW didn't seem to have anything that I thought matched the problem. 5. Is it reasonable to even put a filer's HTTP into an environment that will generate roughly several million hits per day? -- do svidaniya, ~mitch

1 0

converting a disk to be a hot-spare
by Mitch Wright 06 Oct '99

06 Oct '99

Greetings, I've got an F220 running 4.3.3. It is doing fine and we've recently moved all of it's workload off to another filer. This filer had all of the disks in use for data (sans the one for parity of course), but what I'd like to do is take back a disk as a hot-spare. There is no longer any data on the filer that we care about with the exeception of the OS itself. Is there a way I can reclaim one of the disks as a hot-spare without reinstalling from scratch? -- do svidaniya, ~mitch

1 0

Re: Backup restore times
by Darrell Fuhriman 06 Oct '99

06 Oct '99

Dave Heiland <dheiland(a)lucent.com> writes: > entire filer. Thats just about ok, but the problem is that restores of user > files can also take up to 8 hours which is a bit long. The problem may get even It's much, much too long in fact. One has the option of making smaller volumes, but that's not always the best option. The problem is *greatly* aggravated if one has to use incrementals also. One thing that would really, really help is if the NDMP v3 spec were (finally) finalized and implemented. v3 includes the ability to send tape positioning information to the backup system, meaning restores can quickly find the correct tape file & record for a given disk file. I don't like the fact that the positioning information is defined on an implementation specific basis, though. After that, netapp needs to either modify dump, or create another backup utility that can actually make use of said information. Disk capacity is rapidly out-pacing tape capacity and speed -- an issue Netapp needs to address sooner rather than later. Doing dump on a 400 GB filesystem (6 dlt7000's!) is down right primitive. Darrell

1 0

Shelf firmware
by tkaczma＠gryf.net 05 Oct '99

05 Oct '99

Is there a way to update the shelf firmware? I wondered so I searched NOW, but I couldn't find any meaningfull references. Is there any reason to do so? Tom

1 0

Re: Backup restore times
by Stoltz 05 Oct '99

05 Oct '99

The "substantial time increase" you mention is most likely caused by the creation of the snapshots before each individual backup. With large numbers of directories, this can really add up (as it has for us). We've been waiting on Netapp to fix bug #15998, which will allow us to directly dump the pre-existing snapshots *and* have /etc/dumpdates updated correctly :P You can work around it by manually editting /etc/dumpdates and stripping out the .snapshot portion of "/vol/<volume>/.snapshot/<identifier>" but we were hoping they'd have fixed it by now. Anyway, dumping the pre-existing snapshots speeds things up immensely with large numbers of dump requests. Hope that helps. Bug #15998 TITLE: ndmpd updates the dumpdates table incorrectly when dumping pre-existing snapshots DESCRIPTION: In 5.3, if a dump of a pre-existing snapshot (such as /vol/vol0/.snapshot/hourly.0) is used, the dumpdates table is updated without the ".snapshot/hourly.0". It should only show: /vol/vol0/ 0 Sun Jan 1 00:00:00 1999 However, ndmpd prints out /vol/vol0/.snapshot/hourly.0 0 Sun Jan 1 00:00:00 1999 In message <37FA25BA.82769EDE(a)lucent.com>, Dave Heiland writes: > >We're using Netbackup to backup our F760 filer via NDMP to a directly connecte >d >STK robot. Full backups take about 8 hours - we tell Netbackup to backup the >entire filer. Thats just about ok, but the problem is that restores of user >files can also take up to 8 hours which is a bit long. The problem may get ev >en >worse as we might be moving from 70-80GB of user space to about 200GB, which >would mean any restores would stop the night's backup and take far too long. > >Does anyone have any suggestions as to how to rearrange the backups to reduce >this time?. We've considered backing up individual directories, but the backu >p >time increases substantially. > >Dave >-- >David J. Heiland Lucent Technologies Tel: +44 (0)1666 83-2504 >dheiland(a)lucent.com CIO Department Malmesbury, England

1 0

Backup restore times
by Dave Heiland 05 Oct '99

05 Oct '99

We're using Netbackup to backup our F760 filer via NDMP to a directly connected STK robot. Full backups take about 8 hours - we tell Netbackup to backup the entire filer. Thats just about ok, but the problem is that restores of user files can also take up to 8 hours which is a bit long. The problem may get even worse as we might be moving from 70-80GB of user space to about 200GB, which would mean any restores would stop the night's backup and take far too long. Does anyone have any suggestions as to how to rearrange the backups to reduce this time?. We've considered backing up individual directories, but the backup time increases substantially. Dave -- David J. Heiland Lucent Technologies Tel: +44 (0)1666 83-2504 dheiland(a)lucent.com CIO Department Malmesbury, England

1 0

RE: Two volumes with same name, DOT 5.2.1, how do I fix?
by David Stack 04 Oct '99

04 Oct '99

John, I had this very same problem but do you think I can remember what I did? Of course not. I put a call into Netapp and they cleared it up for me. What's listed in the manual and what I think I did was: vol rename vol0(1) vol1 Your listing does not show the (1) for the foreign volume but I seem to remember that when I did a vol status there was a vol0(1) for the foreign volume. Dave **************************************************************************** ************* EMail: dstack(a)nortelnetworks.com Phone: 403-232-4752 Fax: 781-8850 ESN:765 Addr : Nortel Networks Ltd.; 2924 11 St. NE; Calgary, AB; T2E 7L7;Canada **************************************************************************** ************ > -----Original Message----- > From: John Stoffel [SMTP:john@casc.com] > Sent: Monday, October 04, 1999 8:19 AM > To: toasters(a)mathworks.com > Subject: Two volumes with same name, DOT 5.2.1, how do I fix? > > > Hi guys, > > We're running DataOnTap 5.2.1 on a bunch of F330s. One of them had a > disk failure this weekend, and since I was in for a power shutdown, I > figured I'd just swap in a spare disk from a returned filer (upgrade > to an F740). > > The problem is the before the old F330 was returned, we never did any > of the disk commands to make all the disks into spares. We just > copied the data off and shutdown the system. > > So I took one of these disks from the old Filer and did the usual disk > swap, disk swap procedure to put it into the Filer with the failed > disk. And now comes the problem. The Filer sees the new disk and > scans it in, but hey! It's got volume info on it, and it won't > initialize the disk and just use it on the new filer. > > Sigh... I've tried using the command 'disk fail 8a.4' but it won't let > me because it's part of a volume. > > So I look at the volume status and I get this: > > swtoast> vol status > Volume State Status Options > vol0 online normal root > vol0 offline foreign > partial > > Yup, *two* volumes, both named 'vol0' and I can't get rid of the > foreign one. If I just do 'vol status vol0' I get the online volume > listed. What I want to do is just a volume destroy, or a disk > initialize without any questions asked. But I can't figure out how to > do this. > > Any hints? I do have a call into NetApp, but I've been on hold for > about 15 minutes already... > > Thanks, > John > John Stoffel - Senior Unix Systems Administrator - Lucent Technologies > stoffel(a)lucent.com - http://www.lucent.com - 978-952-7548 > john.stoffel(a)ascend.com - http://www.ascend.com

2 1

Autosupport email changes in 5.3
by Graydon Dodson 04 Oct '99

04 Oct '99

I have recently upgraded a filler to 5.3.2R1P1. This weekend it sent an autosupport email after a successful disk scrub, "ERROR_FREE_DISK_SCRUB". This seems odd to me since nowhere in the autosupport(8) man page is this documented. In fact it explicitly states that autosupport email is ONLY generated for system failures, weekly status and boots from disk. Since we use autosupport email to trigger actions from our administrators does anyone know of a way to suppress this specific autosupport email? Thanks. Graydon Dodson grdodson(a)lexmark.com Lexmark International Inc.

1 0

case #79131, Very Poor Backup performances: 1Gb/H with Veritas Netbackup through NFS
by Pierre Maubert 04 Oct '99

04 Oct '99

Hi, We have several F330 in TIF, 2 of them have an enourmous number of small files: over 3.5M for one single volume. We have increase the maxfiles value to 4328504. We have currently a significative degradation of backup performances on these 2 servers: backup are done at 1.5Gb, on both server, wheras we reach 5Gb/h for the 3 others. Servers are all running Dataontap5.2.3, and are all identical. F330, 128Mb Ram, 24*4Gb disks, I wonder what kind of fine tuning can be done to speed up the backup. -- Regards, Pierre . +------------------------------------------------------+ | Pierre Maubert, Sys_Admin Texas Instruments France | | email:p-maubert@ti.com Tel: +33 49322.2681 | | Fax: +33 49202.4669 Pager: 0606498982 | +------------------------------------------------------+

2 1

Re: Windows Terminal Servers??
by Luke Mewburn 01 Oct '99

01 Oct '99

"Nail, Larry" writes: > Is anyone else using files to store home directories & profiles for Windows > Terminal Servers? Yes. Our NetApps are being used to store our home directories which are shared between: 5 Sun Enterprise servers via NFS 15 SGI workstations via NFS about 100 NT workstations via CIFS 4 NT Terminal Servers via CIFS (~ 125 users) We did have roaming profiles for a while, being saved back to the user's home directory. We've recently converted the Terminal Servers back to mandatory profiles because with ~ 2500 user accounts the local profile store on the Terminal Servers was filling up. Also, some users were complaining that their quotas were filling up with profile junk. Personally, I had no problem with it, but as a sysadmin, I don't have a quota problem ;-) I seem to recall that there might have been a couple of Terminal Server bugs that needed to be worked around (I could ask our NT guys), but there should be sufficient information in TechNet about this.

1 0

Windows Terminal Servers??
by Nail, Larry 01 Oct '99

01 Oct '99

Is anyone else using files to store home directories & profiles for Windows Terminal Servers? ------------------------------------------ Lawrence D. Nail Microsoft Certified Systems Engineer Senior Server Administrator PC Workstation Support Mixed Signal Products http://www1.msp.sc.ti.com/nail/ <http://www1.msp.sc.ti.com/nail/>

1 0

Can anyone send his cache hit rate for his NetCache (+site details) ?
by Eyal Traitel 01 Oct '99

01 Oct '99

1 0

RE: BudTool NDMP backups dying prematurely.
by Melvin, Grant 30 Sep '99

30 Sep '99

Hi Paul, > We were told by NetApp that they are working on a fix, > but the bug hasn't even shown up in the NOW database yet... YMMV. I believe this bug is reported as 17997 - dump fails to handle directory names with a null dos_name. and is now online. Cheers, Grant -----Original Message----- From: ptaylor(a)catalystsg.com [mailto:ptaylor@catalystsg.com] Sent: Wednesday, September 29, 1999 10:59 AM To: Josh Tiefenbach Cc: toasters(a)mathworks.com Subject: Re: BudTool NDMP backups dying prematurely. We had this problem for a while, and we were told to upgrade to DoT 5.3.2D4, and Budtool 4.6.1. This stopped the pauses, but introduced a buffer overflow that makes the netapp dump core whenever it hits a "null DOS condition" while trying to back up. We aren't really sure what a "null DOS condition" is, but it makes us unhappy when we hit it. We were told by NetApp that they are working on a fix, but the bug hasn't even shown up in the NOW database yet... YMMV. Paul Taylor Sr. Systems Engineer Catalyst Solutions Group 215-841-5540 On Wed, 29 Sep 1999, Josh Tiefenbach wrote: > I thought I'd toss this one up for grabs, whilst I try to find support people > to talk to. > > I've got a BudTool 4.6, clustered F760, DOT 5.2.1 setup here, and I'm having > the dickens of a time getting my filers backed up via NDMP. > > Ordinarily I have file history turned on. It would appear that because of > this, I'm ramming head on into NetApp BugID 9378. I do in fact see a large > pause between pass III and pass IV. A 2 hour pause. > > This is a problem, as 2 hours later, the backup dies with the following > message in the logs: > > Command exceeded maximum allowable run time 14400. > > ie, 4 hours. > > Questions I have: > > * Has anybody found a way to circumvent the multiple-hour pause during a > BudTool NDMP backup, _while keeping file history on_ ? > > * Anybody seen the `killing job after 4 hours' message? Anybody know a > workaround? > > While highly sub-optimal, I have a backup running with no file history right > now. I'm concerned that that 4 hour thing is gonna bite me in the butt again. > > Suggestions? Comments? Shotgun shells to rid myself of this headache? > > josh > > -- > This is my .sig! There are many others like it, but this one is mine! >

1 0

Re: NetApps Backup Strategy
by Michael Salmon 30 Sep '99

30 Sep '99

+----- On Wed, 29 Sep 1999 11:42:29 EDT, John Stoffel writes: | | [ Talking about the Legato NDMP module for NetApps ] | | Michael> The module has been released although I haven't gotten it to | Michael> work yet. The installation is pain free at least and the | Michael> server can at least probe the filer although it report 0 | Michael> disks. As the license hasn't arrived yet and we are at | Michael> present between support contracts I haven't complained as | Michael> yet. As you might expect the documentation was minimal, but | Michael> hey if it was easy then anybody could do it ;^). | | I guess this is a step in the right direction, and since I'd love to | move this way can we ask you to keep good notes and let us know what | you have to do to make it work? | | I'm personally running Solstice Backup 5.1 (SUN OEM version) on my | backup servers, and my Toasters are running DOT 5.2.1 right now, so I | realize I'll have to upgrade both sides to make this work, though I'm | hoping that I can get away with only upgrading the toasters. Upgrading the toasters should be enough, you may not need to buy a license in that case as I think that clientpack licensing came in networker 5.5. /Michael

1 0

RE: CIFS, UNIX and ACLs
by Norman, Paul 30 Sep '99

30 Sep '99

Why do I get the feeling that after I answer this question, the next one will be what color underwear I have on? OK, Most of the systems are SUN running Solaris, usually the latest version but some may have older revs. There are probably a few SunOS systems too. We have PCs running Win9X, NT, and Linux. Linux is recent addition but we seem to have quite a few around recently. There is an HP running HP-UX. We have a few SGI boxes but they aren't usually used in testing. Oh ya, there are a number of Dec Alpha systems running Digital UNIX. I don't recall seeing any other gear, but then I don't spend my days traipsing around the labs because I actually have a job in Product Marketing. > -----Original Message----- > From: Eyal Traitel (r55789) [mailto:r55789@email.sps.mot.com] > Sent: Monday, September 27, 1999 2:15 AM > To: Norman, Paul > Cc: 'Graham C. Knight'; toasters(a)mathworks.com > Subject: Re: CIFS, UNIX and ACLs > > > Hi again Paul. > > We would like to ask for what are the testing that you do before a > release is made? > Against which OSs and applications you test it ? > > We feel that this information is somewhat lacking from the NOW... > > > Eyal. > >

4 3

Re: Novell 5/NT 4.0 and Netapp filers
by Luke Mewburn 29 Sep '99

29 Sep '99

"Mike Ball" writes: > We are looking at implementing Network Appliance filers into a shared NT and > Novell environment. I understand as long as "Client for Microsoft > Networking" is loaded on the Novell clients and as long as the Novell users > are being authenticated by the NT domain when they cross over to access the > filer, the Novell users will be able to store/retrieve their files on the > Netapp. My question is has anyone using Novell 5 in this environment ran > into any issues with the filer? Our department is primarily an NFS shop with an NT domain for the PCs running NT Workstation or NT Terminal Server that want to access the same home directories via CIFS. This works extremely well for us. However, our University is a Novell shop, so some of our users have to have access to Netware volumes in other departments or faculties. What we've discovered is (some of this is from slightly hazy memory; I'm the UNIX and filer admin, I only use PCs occassionally): * you should really have an NT domain to authenticate the CIFS sessions. As long as the users have the Netware client configured to synchronise passwords things should be ok on that front (log in once, and one password gets you to your NT and Netware drives) * depending on which system type you access more frequently (NT or Netware objects), you may want to tweak the `Network Services Order' (or whatever it's called) in the Control Panels -> Network section. Because our PCs are primarily accessing CIFS drives, we move Microsoft's requestor above the Netware one. We found that browsing objects through Network Neighbourhood that are of the `lower priority' type can take a while, especially if the primary `tree' is big and/or slow. * I've found the Netware client to be slow and buggy, when compared to the standard NT domain client. Due to (misguided, IMHO) university policies about `single sign-on through the central Netware NDS', we will be trialling the use of NDS for NT as a replacement for our NT domain; this may (assuming it works) make it easier for users with respect to password/account details. That would mean we wouldn't have to install the Netware client on most machines either (see above). Luke.

1 0

RE: File locking, sendmail and toasters
by Mohler, Jeff 29 Sep '99

29 Sep '99

I'll hit the WWW on that, a little learnin never hurt *grin* Thanks. -----Original Message----- From: Nick Christenson [mailto:npc@sendmail.com] Sent: Wednesday, September 29, 1999 1:33 PM To: jeff.mohler(a)netapp.com Cc: toasters(a)mathworks.com Subject: Re: File locking, sendmail and toasters > Could you not use ProcMail for local delivery (dot locking)? > > No..im not a NetApp engineer, but in past lives that's what Ive done before when > presented with mailspools via NFS. In a general sense with general NFS implementations, that's not enough by itself to be sure you're safe. You certainly want to use file locking rather than lockd, but using files isn't a sufficient condition to be sure that your mail won't get messed up. Note, though, that you can "get away" with doing a naive NFS based mail implementation and statistically never see problems, but there is usually a race condition in there somewhere. -- Nick Christenson npc(a)sendmail.com

1 0

RE: File locking, sendmail and toasters
by Mohler, Jeff 29 Sep '99

29 Sep '99

Could you not use ProcMail for local delivery (dot locking)? No..im not a NetApp engineer, but in past lives that's what Ive done before when presented with mailspools via NFS. -----Original Message----- From: Nick Christenson [mailto:npc@sendmail.com] Sent: Wednesday, September 29, 1999 1:06 PM Cc: toasters(a)mathworks.com Subject: Re: File locking, sendmail and toasters > Nick Hilliard (nick(a)iol.ie) said, on [990929 13:54]: > > > > If you need to have text-based NFS-mounted client mailboxes, the only > > reliable option is to use a mailbox format which doesn't require locking, > > such as Maildir. > > Amen. > The only thing I've ever seen that didn't immediately give me the willies is: > > http://www.earthlink.com/about/papers/mailarch.html Just for reference, I was lead author of that paper. Running mail over NFS is doable if you're really, really careful. You will definitely want to replace the rpc.lockd locking mechanism with something better. This means modifying every program anyone uses to touch the mailbox (have to do that if you go with maildir as well). If I understand your system and requirements, which is in doubt, I don't think this is a solution I would recommend for your situation. Hope this helps. -- Nick Christenson npc(a)sendmail.com

2 1

File locking, sendmail and toasters
by Stephen C Woods 29 Sep '99

29 Sep '99

We have a medium sized environment with several discrete mail domains. The mailservers for the domains are (for historical reasons) all of a different type (HP-UX, Solaris, AIX and SGI) but are all running sendmail-8.8.8. Now in the documentation from sendmail they specificly recomend AGAINST running the mail server on a machine that is an NFS client for the user mailbox filesystem. Is anyone out there actually doing this? With a ~500 workstation, 5 mailserver environment (we'll still have multiple different mail directories one for each mail domain)? With a mixed (HP, SUN, AIX, SGI, NEXT, Linux) environment? <scw> postmaster(a)seas.ucla.edu ----- Stephen C. Woods; UCLA SEASnet; 2567 Boelter hall; LA CA 90095; (310)-825-8614 Finger for public key scw(a)cirrus.seas.ucla.edu,Internet mail:scw@SEAS.UCLA.EDU

5 4

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

toasters