toasters

toasters@lists.teaparty.net

13515 discussions

RE: solaris 2.6 nfs client tuning?
by Venier, Federico 19 Jan '01

19 Jan '01

I have seen this before. It has something to do with the mmap routine called by the cp program in Solaris 2.6. I believe that it is fixed in Solaris 2.8 build 33. Using a dd everyting goes OK. Maybe it is also possible to "tar" the directory. CIAO =============================================================== Federico VENIER System Engineer Phone : +39 039 6858483 Network Appliance Srl Fax : +39 039 6858485 Centro Direzionale Torri Bianche Mob. : +39 0348 4719025 Palazzo Larice mail : fvenier(a)netapp.com 20059 Vimercate (MI) - Italy http://www.netapp.com =============================================================== -----Original Message----- From: Toal, Dave [mailto:dave.toal@t-t.com] Sent: lunedì 15 gennaio 2001 21:00 To: 'toasters(a)mathworks.com' Cc: Toal, Dave Subject: solaris 2.6 nfs client tuning? Hey, all. This lots-of-nfs-mounts topic inspires me to ask yet another sanity-check question: Currently I'm fixing a process which moves lots of data from one vol to another. Can't simply vol copy because there are things I can't over-write on the destination. This data is in about 40 directories, with maybe 2G in each. The method I'm using is to set up many pairs of mounts, one for each directory, all from the same two vols. This is because the nfs client, a sun 450 running solaris 2.6, shows huge lag for ls, cp, whatever, in a mounted directory where a cp is already running. iostat shows the nfsd devices at 100% busy -- for destination and source -- when a cp runs between a mount-point pair. So: filer1:/vol/A/d1 /mnt/A_01 filer2:/vol/B/d1 /mnt/B_01 filer1:/vol/A/d2 /mnt/A_02 filer2:/vol/B/d2 /mnt/B_02 filer1:/vol/A/d3 /mnt/A_03 filer2:/vol/B/d3 /mnt/B_03 ad nauseum... and if I cp /mnt/A_01/fatfile to /mnt/B_01 then there's no i/o lag whatsoever in /mnt/A_02. These filers are 760's, by the way. Then I discovered a serious "sweet spot" at a very low number. "max" is the number of simultaneous copies between mount point pairs. "k/sec" are my visual estimates of _each_ cp rate, from watching iostat for about a minute; rates jump around by +- 100 k/sec. Column on the right is total throughput for all cp. max=3 ~ 2300 k/sec |= 6900 max=4 ~ 1800 k/sec |= 7200 max=5 ~ 1500 k/sec |= 7500 max=6 ~ 1200 k/sec |= 7200 max=8 ~ 800 k/sec |= 6400 max=9 ~ 800 k/sec |= 7200 max=10 ~ 600 k/sec |= 6000 I've seen much higher data rates from the 760's. So I'm thinking this means I need to learn nfs client tuning for solaris. Yes? Dave -- dave toal systems optimist Thomson & Thomson N. Quincy, MA -----Original Message----- From: Jeffrey Krueger To: Rainchik, Aleksandr (MED, Non GE) Cc: 'Jeffrey Krueger '; 'toasters(a)mathworks.com ' Sent: 1/15/01 1:37 PM Subject: Re: /home layout with many filers and NIS automount On Sat, Jan 13, 2001 at 03:16:20PM -0600, Rainchik, Aleksandr (MED, Non GE) wrote: > > There is another thing against > > #auto_home NIS map > > user1 filer2:/vol/vol1/& > design. We have a lot of development servers. So any time you > add/delete/modify a single user you should go trough _all_ the > NIS/automount clients and run "automount" to let automountd know > that passwd map has been just updated. Am I wrong again? This is correct, but I simple cron job which runs periodically on all NIS clients should make short work of this task. =) -- Jeff

1 0

RE: Mark Conley's comments on used NetApp systems
by ScottUCE＠aol.com 19 Jan '01

19 Jan '01

I am responding to a posting last week from Mark Conley, a Network Appliance salesman from Northern California, regarding used NetApp systems. Mark wrote "FYI on the used equipment, none of the software licenses on Netapp equipment are transferable. Which means you are paying for hardware only!" Mark Conley does not make policy for Network Appliance. In fact, Union Computer has been involved in the sale or brokering of numerous NetApp systems and parts, and, in every case, Network Appliance has been a pleasure to work with; they have accepted the equipment for maintenance service and have, where applicable, allowed software licenses to transfer without charge. The market for used NetApp gear is growing, and Network Appliance should enunciate its policy with respect to the re-sale of its used systems; in particular, their boilerplate software license agreement clearly gives Network Appliance the discretion to allow or forbid the seller's transfer of software licenses to buyers of used equipment. Without doubt, it is in the clear interest of Network Appliance's customers that they be able to re-sell their systems without harassment. We are currently working with several NetApp users in the midst of severe business downturns, forcing them to put their systems on the market. If Network Appliance were to interfere with the transfer of their systems' NFS and CIFS licenses, the value of their F760 heads would be reduced to that of spare parts kits. Unless and until Network Appliance clarifies its policy on the re-sale of its systems, NetApp's end-user and/or leasing company customers -- before they buy their next box -- should request a modification of their software license agreements so as to limit Network Appliance's discretion to forbid software license transfer, thus helping to safeguard the value of their investments in their NetApp systems. Scott Fischmann Union Computer Exchange, Inc. 6233 Idylwood Lane Minneapolis, MN 55436 Phone: 952 935 7282 www.unioncomputer.com

3 2

RE: Backups of both NFS and CIFS Directories
by Benn, Paul 18 Jan '01

18 Jan '01

Glenn, To retain both permission styles you'll need to backup using the local dump command, an NDMP enabled backup application or backup over CIFS. Backing up over NFS will not retain NTFS file permissions or DOS attributes. If you back up over CIFS, not only are the UNIX and NT file permissions retained but we also backup the DOS file attributes and track an archive bit for files with UNIX-style permissions. If backing up over CIFS is your plan, have a look at our white paper on this subject: http://www.netapp.com/tech_library/3052.html Regards, Paul Benn Network Appliance -----Original Message----- From: Glenn Case - Informix Software [mailto:glennc@informix.com] Sent: Thursday, January 18, 2001 2:57 PM To: toasters(a)mathworks.com Subject: Backups of both NFS and CIFS Directories I have a new netapp that I am in the process of trying to set up It will be supporting both NFS and CIFS Clients. I was wondering if anyone knew how Backing up the Permissions for the Two different permissions styles works with the different backup styles ? I am concerned that If I back the machine up "Wrong" we would lose the permissions of CIFS users files. For example If I back the machine up with tar from a UNIX machine I know the permissions will be lost. If I back it up with the built in backup command and then restore with solaris ufsrestore It will also lose the CIFS permissins. So what do I need to do to make sure that the CIFS permissions are saved ? If this is a FAQ question or I just need to RTFM Please just refer me to the appropreate document. (With URL if avalable ) Thanks in advance Glenn

1 0

Backups of both NFS and CIFS Directories
by Glenn Case - Informix Software 18 Jan '01

18 Jan '01

I have a new netapp that I am in the process of trying to set up It will be supporting both NFS and CIFS Clients. I was wondering if anyone knew how Backing up the Permissions for the Two different permissions styles works with the different backup styles ? I am concerned that If I back the machine up "Wrong" we would lose the permissions of CIFS users files. For example If I back the machine up with tar from a UNIX machine I know the permissions will be lost. If I back it up with the built in backup command and then restore with solaris ufsrestore It will also lose the CIFS permissins. So what do I need to do to make sure that the CIFS permissions are saved ? If this is a FAQ question or I just need to RTFM Please just refer me to the appropreate document. (With URL if avalable ) Thanks in advance Glenn

1 0

RE: FW: Multicast support in ONTAP?
by Rainchik, Aleksandr (MED, Non GE) 18 Jan '01

18 Jan '01

MU3 for Solaris 7? There is only MU1 available for Solaris 8 at this time, at least at sunsolve.sun.com. -----Original Message----- From: Chris Thompson [mailto:cet1@cus.cam.ac.uk] Sent: Wednesday, January 17, 2001 6:29 PM To: Rainchik, Aleksandr (MED, Non GE) Cc: toasters(a)mathworks.com Subject: Re: FW: Multicast support in ONTAP? Aleksandr.Rainchik(a)med.ge.com (Rainchik Aleksandr) writes [...] > Can I ask you about that "Solaris's IP network multipathing > failover and fallback" you've mentioned on toasters mailing > list: > > Is is something that Solaris 8 has right out of box? Yes, if you use MU3 or later. Or you can apply a raft of patches to an earlier version (109898-01 109742-02 109279-05 109900-01 109472-03 109740-02 109902-01 109904-01 109906-01); but this won't get the man pages updated (cue for a rant about documentation maintenance, but this isn't the place for it). Chris Thompson University of Cambridge Computing Service, Email: cet1(a)ucs.cam.ac.uk New Museums Site, Cambridge CB2 3QG, Phone: +44 1223 334715 United Kingdom.

2 1

Re: Mark Conley's comments on used NetApp systems
by mds＠gbnet.net 18 Jan '01

18 Jan '01

I'd just like to pitch in and say that a healthy used NetApp market would be a fine thing indeed for those of us who would like to buy the equipment but can't normally afford it. If the software licenses transferred, NetApp would still have made their main money from the original sale. Support, however, *should* be a different ball-game, and people shouldn't be able to just transfer the client-NetApp relationship with the hardware automatically and purchasers of 2nd-hand hardware shouldn't expect to get support from NetApp without paying for it. NetApp could, of course, offer support tiers, allowing for software updates separate from hardware RMA, separate from SE on-site attendance. This would allow the budget-limited and the 2nd-hand market to benefit and grow, while not missing out on the support bucks that people would be willing to spend with them. Support can be a viable revenue stream and not just something that you do to keep selling hardware. Just look at the Enhanced Margin Company - they get you coming and going! I've observed before (but possibly not here) that products in NetApp's line get EOL'd rather more aggressively than other maufacturers products (Sun, IBM, HP, even SGI all sell/support old models longer). Having a viable re-sale market could also take the sting out of that policy. I guess it all comes down to whether it could be seen to be made to benefit NetApp's bottom-line and medium-long term plans or not. I'd like to think it could be. -- -Mark ... an Englishman in London ...

1 0

Add more disks and upgrade to 6.0 - which first?
by Chris Thompson 18 Jan '01

18 Jan '01

There are two things I want to do within the next three months or so, neither of them urgent: 1. Add 2 x 18GB discs (currently spare) to a volume that currently has 6 x 18 GB data discs. 2. Upgrade the F740 from ONTAP 5.3.7R? to 6.0.1R?. Is there any advantage to be gained by doing these in any particular order? Chris Thompson University of Cambridge Computing Service, Email: cet1(a)ucs.cam.ac.uk New Museums Site, Cambridge CB2 3QG, Phone: +44 1223 334715 United Kingdom.

2 1

FW: Multicast support in ONTAP?
by Rainchik, Aleksandr (MED, Non GE) 18 Jan '01

18 Jan '01

Hello, Can I ask you about that "Solaris's IP network multipathing failover and fallback" you've mentioned on toasters mailing list: Is is something that Solaris 8 has right out of box? Or you need to purchase additional pieces of software, like Sun Trunking or Sun Cluster Services, to implement such a feature? We are mostly Solaris 7 shop right now, but we really need this feature, so I just wonder if moving to 8 will do it. TIA -- Aleksandr Rainchik, Sun CSA -----Original Message----- From: Brian L. Brush [mailto:brian@paradyne.com] Sent: Tuesday, January 16, 2001 1:48 PM To: toasters(a)mathworks.com Subject: Multicast support in ONTAP? I'm experimenting with back-to-back (crossover) links between Sun boxes and filers. Specifically, I'm trying to use Solaris 8's IP network multipathing failover and fallback with such connections. The problem is that Solaris is unable to recognize the failure of a NIC directly connected to a filer. In testing, the feature works as expected on our normal switched network. Solaris's IP network multipathing failover and fallback (what a mouthful!) works by initially sending multicast ICMP echo requests (to ALL-SYSTEMS.MCAST.net). The Solaris box relies on routers and hosts to respond to the multicast requests. Then it sends periodic unicast ICMP echo requests to those specific routers/hosts and knows a NIC has failed when the responses stop. Snoop on the Solaris box reveals the filer is not responding to the multicast requests. The Solaris box, therefore, never even begins its monitoring because it thinks there is nothing to ping. Here is an excerpt: 192.168.0.5 -> ALL-SYSTEMS.MCAST.net ICMP Echo request (ID: 8194 Sequence number: 1) 192.168.0.5 -> ALL-SYSTEMS.MCAST.net ICMP Echo request (ID: 8194 Sequence number: 1) 192.168.0.5 -> ALL-SYSTEMS.MCAST.net ICMP Echo request (ID: 8194 Sequence number: 1) The filer is definitely sitting on the other end of the cable: ping -s 192.168.0.1 PING 192.168.0.1: 56 data bytes 64 bytes from 192.168.0.1: icmp_seq=0. time=1. ms 64 bytes from 192.168.0.1: icmp_seq=1. time=0. ms 64 bytes from 192.168.0.1: icmp_seq=2. time=0. ms 64 bytes from 192.168.0.1: icmp_seq=3. time=0. ms Solaris box is an Ultra 1 with Solaris 8 10/00. Filer is an F760 running Data ONTAP 5.3.7. We've tested Sun ge to NetApp gigabit and also Sun hme to NetApp quad fast ethernet. I'm interested in hearing any suggestions that might make this configuration work. --Brian L. Brush Senior Systems Administrator Paradyne Corporation

4 3

RE: 6.0.1R1 Filerview -- How to print
by Darragh, Stephen J 17 Jan '01

17 Jan '01

Found the answer right after I sent the request. You have to open the frame in a new browser window so you can print. Sorry Stephen Darragh > -----Original Message----- > From: Darragh, Stephen J [SMTP:stephen.j.darragh@lmco.com] > Sent: Wednesday, January 17, 2001 8:39 AM > To: 'toasters(a)mathworks.com' > Subject: 6.0.1R1 Filerview -- How to print > > I hope this is one of those questions that I will say "of course, why > didn't > I see that". > > 6.0.1R1 filerview running from netscape 4.7.5 on a sun solaris 8 system. > > There are now a great number of reports and disk information, but I can't > see how to print them. > There is no CTRL-P, or alt-click to bring up a print option. > > I could use xwd, but there must be a simple way that I am just missing. > > Thanks in advance... > > Stephen Darragh

1 0

RE: /home layout with many filers and NIS automount
by Rainchik, Aleksandr (MED, Non GE) 17 Jan '01

17 Jan '01

There is another thing against > #auto_home NIS map > user1 filer2:/vol/vol1/& design. We have a lot of development servers. So any time you add/delete/modify a single user you should go trough _all_ the NIS/automount clients and run "automount" to let automountd know that passwd map has been just updated. Am I wrong again? Thank you. P.S. I know about rsh/ssh :) -----Original Message----- From: Jeffrey Krueger To: Rainchik, Aleksandr (MED, Non GE) Cc: toasters(a)mathworks.com Sent: 1/12/01 2:30 PM Subject: Re: /home layout with many filers and NIS automount On Fri, Jan 12, 2001 at 10:47:54AM -0600, Rainchik, Aleksandr (MED, Non GE) wrote: > Hello, > > I just need a word of advice on how to structure /home if you > use automount, have 5-10 home directory servers (NetApp filers) > and around 1000 users. Point of clarification: 5-10 filers for 1,000 users? What model filers are they and I hope they are serving other data besides homedirs! At one point we had about 15,000 home directories mostly on two NetApps - a F330 and a F540 - and they were pretty bored filers. One way to simplify your design would be to reduce the number of filers serving homedirs, especially for 1,000 users. Two good reasons to spread out homedirs are for bandwidth constraints and to spread risk/reduce points of failure. But if you have consistently high bandwidth between all users and filers than 2-4 homedir servers should be more than plenty mitigation of risk considering the cost of additional filers versus the additional benefit (considering NetApps impressive reliability). > I'm personally is looking for /home/homeserver-vol#/user-name > scheme, so my automount maps will be nice, small and manageble: > > #auto_home NIS map > filer1-0 filer1:/vol/vol0 > filer1-1 filer1:/vol/vol1 > ... > filer5-5 filer5:/vol/vol5 This is a very restrictive design because it makes homedir moves extremely conspicuous. It also flies a bit in the face of a large selling point of the automounter - eliminating host-based network paths. If a user is moved from one filer to another, or even from one volume to another, they will have a whole new path to their home directory. Worse yet, any scripts or application configurations which use the full path to their home directory will need to be re-coded or re-configured - YUK! Being able to "hide" the location of a home directory (or any other remote filesystem) is extremely useful. > But it seems like users prefer /home/username scheme and it leads > to 1000+ NIS automount map for /home, like > > #auto_home NIS map > user1 filer2:/vol/vol1/& > .... > user9999 filer3:/vol/vol4/& This is the way we do it, and the way I'd venture to guess most/many places do it. From the user perspective, it is very simple. From the admin perspective, it is easy to manage. We've found it to give maximum flexibility with minimum hassle. > Do you know if there is more elegant solution for this problem? Can you > tell me, what is your preference for /home layout? What do you mean by elegant? The above is pretty darned elegant, and for 1,000 users it should be easily manageable too. We currently have 10,223 home directories and we've got one big home directory map like above. It works great! One recommendation is you may want to use a script or database to help generate that automount map. The easiest way would be to write a shell/perl script that scanned all your "homedir" directories on each filer, ensured username uniqueness and then build the automount map. For added flexibility, we use a small database. Each row in the homedir table has the username (mountpoint under /home), the filer, and the path to their homedir. At some interval a script comes thru and builds the map. Hope this info helps! -- Jeff -- ------------------------------------------------------------------------ ---- Jeff Krueger, NetApp CA E-Mail: jeff(a)qualcomm.com Senior Engineer Phone: 858-651-6709 NetApp Filers / UNIX Infrastructure Fax: 858-651-6627 QUALCOMM, Inc. IT Engineering Web: www.qualcomm.com

4 5

6.0.1R1 Filerview -- How to print
by Darragh, Stephen J 17 Jan '01

17 Jan '01

I hope this is one of those questions that I will say "of course, why didn't I see that". 6.0.1R1 filerview running from netscape 4.7.5 on a sun solaris 8 system. There are now a great number of reports and disk information, but I can't see how to print them. There is no CTRL-P, or alt-click to bring up a print option. I could use xwd, but there must be a simple way that I am just missing. Thanks in advance... Stephen Darragh

1 0

ADV: Search Engine Registration
by mb6161＠arabia.com 17 Jan '01

17 Jan '01

Removal instructions below. I saw your listing on the internet. I work for a company that specializes in getting clients web sites listed as close to the top of the major search engines as possible. Our fee is only $29.95 per month to submit your site at least twice a month to over 350 search engines and directories. To get started and put your web site in the fast lane, call our toll free number below. Mike Bender 888-532-8842 To be removed call: 888-800-6339 X1377

1 0

ENTERPRISE EMAIL
by Joe Luchtenberg 17 Jan '01

17 Jan '01

There are a number of good threads in the toasters archives touching on large-scale e-mail implementation. My question is what other sites/newsgroups would you suggest are good primary resources for information/threads on enterprise e-mail? (I already have the Earthlink whitepaper from an earlier toasters thread) FWIW, the particular implementation under consideration is open to most anything at this time -- Exchange, sendmail, web-based, POP, etc. TIA. Joe Joe Luchtenberg Dataline, Inc. New number and email, please update: 757.858.0600 858.0606 fax joe.luchtenberg(a)data-line.com

3 3

Multicast support in ONTAP?
by Brian L. Brush 17 Jan '01

17 Jan '01

I'm experimenting with back-to-back (crossover) links between Sun boxes and filers. Specifically, I'm trying to use Solaris 8's IP network multipathing failover and fallback with such connections. The problem is that Solaris is unable to recognize the failure of a NIC directly connected to a filer. In testing, the feature works as expected on our normal switched network. Solaris's IP network multipathing failover and fallback (what a mouthful!) works by initially sending multicast ICMP echo requests (to ALL-SYSTEMS.MCAST.net). The Solaris box relies on routers and hosts to respond to the multicast requests. Then it sends periodic unicast ICMP echo requests to those specific routers/hosts and knows a NIC has failed when the responses stop. Snoop on the Solaris box reveals the filer is not responding to the multicast requests. The Solaris box, therefore, never even begins its monitoring because it thinks there is nothing to ping. Here is an excerpt: 192.168.0.5 -> ALL-SYSTEMS.MCAST.net ICMP Echo request (ID: 8194 Sequence number: 1) 192.168.0.5 -> ALL-SYSTEMS.MCAST.net ICMP Echo request (ID: 8194 Sequence number: 1) 192.168.0.5 -> ALL-SYSTEMS.MCAST.net ICMP Echo request (ID: 8194 Sequence number: 1) The filer is definitely sitting on the other end of the cable: ping -s 192.168.0.1 PING 192.168.0.1: 56 data bytes 64 bytes from 192.168.0.1: icmp_seq=0. time=1. ms 64 bytes from 192.168.0.1: icmp_seq=1. time=0. ms 64 bytes from 192.168.0.1: icmp_seq=2. time=0. ms 64 bytes from 192.168.0.1: icmp_seq=3. time=0. ms Solaris box is an Ultra 1 with Solaris 8 10/00. Filer is an F760 running Data ONTAP 5.3.7. We've tested Sun ge to NetApp gigabit and also Sun hme to NetApp quad fast ethernet. I'm interested in hearing any suggestions that might make this configuration work. --Brian L. Brush Senior Systems Administrator Paradyne Corporation

2 1

RE: /home layout with many filers and NIS automount
by Premanshu Jain 16 Jan '01

16 Jan '01

Anybody using NDMP for NT. I need to backup NT servers on a P3000 which supports only NDMP. Where can we get NDMP for NT? prem Premanshu Jain Nortel Networks, Content Networks(Shasta) 2305, Mission College Blvd, Santa Clara, CA 95054 Direct: (408) 565-3573 ESN: 655-3573 eMail: prjain(a)shastanets.com Web: http://www.nortelnetworks.com/ipservices -----Original Message----- From: Rob Windsor [mailto:windsor@adc.com] Sent: Monday, January 15, 2001 12:45 PM To: toasters(a)mathworks.com Subject: Re: /home layout with many filers and NIS automount On Mon, 15 Jan 2001 10:37:38 PST, Jeffrey Krueger wrote: >> There is another thing against >> > #auto_home NIS map >> > user1 filer2:/vol/vol1/& >> design. We have a lot of development servers. So any time you >> add/delete/modify a single user you should go trough _all_ the >> NIS/automount clients and run "automount" to let automountd know >> that passwd map has been just updated. Am I wrong again? > This is correct, but I simple cron job which runs periodically on all NIS > clients should make short work of this task. =) Er? Solaris? That's one of the (few) beauties of NIS -- you don't have to go around telling everything that a map has changed. I've never gone machine-to-machine running automount(1) to kick it in the pants just because we changed the password or auto_home map. The _worst_ is that ncsd caches something and we need to kill and restart it. Rob++ ---------------------------------------- Rob Windsor E-Mail - mailto:rob_windsor@adc.com Senior Unix Systems Administrator Voice - phone:972-680-6919 Computer Services Fax - phone:972-680-0370 Broadband Access and Transport Group __o ADC Telecommunications _`\<,_ Richardson, TX 75082 (_)/ (_)

1 0

RE: /home layout with many filers and NIS automount
by Darragh, Stephen J 16 Jan '01

16 Jan '01

There are two types of automount maps (direct and indirect). I believe that direct maps are the ones that need to be signaled when a map changes. We have had to use direct maps for nested filesystem mounts, but do this only on an exception basis. See the O'Reilly book "Managing NFS and NIS" by Hal Stern. Hope this helps Stephen Darragh > -----Original Message----- > From: Rob Windsor [SMTP:windsor@adc.com] > Sent: Monday, January 15, 2001 3:45 PM > To: toasters(a)mathworks.com > Subject: Re: /home layout with many filers and NIS automount > > On Mon, 15 Jan 2001 10:37:38 PST, Jeffrey Krueger wrote: > > >> There is another thing against > >> > #auto_home NIS map > >> > user1 filer2:/vol/vol1/& > >> design. We have a lot of development servers. So any time you > >> add/delete/modify a single user you should go trough _all_ the > >> NIS/automount clients and run "automount" to let automountd know > >> that passwd map has been just updated. Am I wrong again? > > > This is correct, but I simple cron job which runs periodically on all > NIS > > clients should make short work of this task. =) > > Er? Solaris? That's one of the (few) beauties of NIS -- you don't have > to go around telling everything that a map has changed. > > I've never gone machine-to-machine running automount(1) to kick it in the > pants just because we changed the password or auto_home map. The _worst_ > is that ncsd caches something and we need to kill and restart it. > > Rob++ > ---------------------------------------- > Rob Windsor E-Mail - mailto:rob_windsor@adc.com > Senior Unix Systems Administrator Voice - phone:972-680-6919 > Computer Services Fax - phone:972-680-0370 > Broadband Access and Transport Group __o > ADC Telecommunications _`\<,_ > Richardson, TX 75082 (_)/ (_) >

1 0

inode limitation?
by Magnus Thorne 16 Jan '01

16 Jan '01

Greetings, I'm planning on using a NetApp 760 to store *many* small files. However, I'm getting vague answers from NetApp for my problem, so I figure I'd ask here. NetApp has told me that they don't recommend that I have more than 250,000 inodes per volume. Is this a real limitation, because my NetApps currently have 6.1 million inodes available? The online documentation says I only need to add more hard drives to get more inodes (1 per 32K). Has anyone used the netapp for small files? Small means millions of < 60K. Also, I need all these files available on one share. Also, what should I look out for? I know to disable the read ahead cache. I plan to not use snapshots. thanx, -Mag

4 3

FW: /home layout with many filers and NIS automount
by Rainchik, Aleksandr (MED, Non GE) 15 Jan '01

15 Jan '01

Basically, if it's indirect map, as auto_home, you don't need to kick "automount" on every client, but if it's direct map - you have to. -----Original Message----- From: Jeffrey Krueger [mailto:jkrueger@qualcomm.com] Sent: Monday, January 15, 2001 12:38 PM To: Rainchik, Aleksandr (MED, Non GE) Cc: 'Jeffrey Krueger '; 'toasters(a)mathworks.com ' Subject: Re: /home layout with many filers and NIS automount On Sat, Jan 13, 2001 at 03:16:20PM -0600, Rainchik, Aleksandr (MED, Non GE) wrote: > > There is another thing against > > #auto_home NIS map > > user1 filer2:/vol/vol1/& > design. We have a lot of development servers. So any time you > add/delete/modify a single user you should go trough _all_ the > NIS/automount clients and run "automount" to let automountd know > that passwd map has been just updated. Am I wrong again? This is correct, but I simple cron job which runs periodically on all NIS clients should make short work of this task. =) -- Jeff

1 0

Re: inode limitation?
by Ketil Kirkerud Elgethun 15 Jan '01

15 Jan '01

Magnus Thorne <Magnus.Thorne(a)evoice.com> writes: > NetApp has told me that they don't recommend that I have more than 250,000 > inodes per volume. Are you sure you're not confusing this with "per directory" ? Even though WAFL is fairly efficient when it comes to large directories, there are limits to what you would want to experience.. > Is this a real limitation, because my NetApps currently > have 6.1 million inodes available? Neo, it's not. This is a 760: df Filesystem kbytes used avail capacity Mounted on /vol/vol0/ 343184120 265805356 77378764 77% /vol/vol0/ df -i Filesystem iused ifree %iused Mounted on /vol/vol0/ 13987164 6012814 70% /vol/vol0/ And it's running just fine (mailspool with one file per message) > The online documentation says I only > need to add more hard drives to get more inodes (1 per 32K). You can have 1 inode per 4K, IIRC. > Has anyone > used the netapp for small files? Small means millions of < 60K. Also, I > need all these files available on one share. As you can see: yes, no problems. > > Also, what should I look out for? I know to disable the read ahead cache. I'm not sure that would help a lot with small files - this is more of a win wit large files (or possible very slow clients) > I plan to not use snapshots. Whether or not to use snapshots is more a question of the change rate in your data set than the number of files - the 760 shown above is running with readahead set, and snapshots. No problems whatsoever. Regards, -- ---Ketil Kirkerud Elgethun, Infostream ASP

1 0

RE: inode limitation?
by Toal, Dave 15 Jan '01

15 Jan '01

Jim, The website I fix uses very small files to handle details of customer events so they can be billed. One file per event. I'm not saying this is a good design. <chuckle> smith> maxfiles spool Volume spool: maximum number of files is currently 8258829 (1428275 used). smith> df -i /vol/spool Filesystem iused ifree %iused Mounted on /vol/spool/ 1428275 6830554 17% /vol/spool/ smith> df /vol/spool Filesystem kbytes used avail capacity Mounted on /vol/spool/ 46944924 14936040 32008884 32% /vol/spool/ /vol/spool/.snapshot 0 7834048 0 ---% /vol/spool/.snapshot smith> How come it's not "maxfiles /vol/spool" -- different coders? Dave -----Original Message----- From: Jim VandeVegt To: Magnus Thorne; 'toasters(a)mathworks.com' Sent: 1/15/01 2:38 PM Subject: Re: inode limitation? --- Magnus Thorne <Magnus.Thorne(a)evoice.com> wrote: > NetApp has told me that they don't recommend that I have more than > 250,000 > inodes per volume. Is this a real limitation, because my NetApps > currently > have 6.1 million inodes available? The online documentation says I > only > need to add more hard drives to get more inodes (1 per 32K). Has > anyone > used the netapp for small files? Small means millions of < 60K. > Also, I > need all these files available on one share. The training I received from NetApp personnel indicated that a particular strength of WAFL is the ability to add more inodes on the fly. There wasn't any mention of any particular recommended limit. A quick check of my system shows that even my smallest volumes currently have "Maximum Numer of Files" set over 1 million. On a 760, one volume with 330 GB of data space, 13 million on MAX files, 550 thousand currently. Another with 150 GB of data space, 6 million on MAX files, 414 thousand currently. As far as I know, no one at our site purposely expanded any of the MAX settings. ===== Jim VandeVegt vandevegt(a)yahoo.com vandevegt(a)avaya.com http://members.xoom.com/vandevegt Toilet paper should come off the front! __________________________________________________ Do You Yahoo!? Get email at your own domain with Yahoo! Mail. http://personal.mail.yahoo.com/

1 0

solaris 2.6 nfs client tuning?
by Toal, Dave 15 Jan '01

15 Jan '01

Hey, all. This lots-of-nfs-mounts topic inspires me to ask yet another sanity-check question: Currently I'm fixing a process which moves lots of data from one vol to another. Can't simply vol copy because there are things I can't over-write on the destination. This data is in about 40 directories, with maybe 2G in each. The method I'm using is to set up many pairs of mounts, one for each directory, all from the same two vols. This is because the nfs client, a sun 450 running solaris 2.6, shows huge lag for ls, cp, whatever, in a mounted directory where a cp is already running. iostat shows the nfsd devices at 100% busy -- for destination and source -- when a cp runs between a mount-point pair. So: filer1:/vol/A/d1 /mnt/A_01 filer2:/vol/B/d1 /mnt/B_01 filer1:/vol/A/d2 /mnt/A_02 filer2:/vol/B/d2 /mnt/B_02 filer1:/vol/A/d3 /mnt/A_03 filer2:/vol/B/d3 /mnt/B_03 ad nauseum... and if I cp /mnt/A_01/fatfile to /mnt/B_01 then there's no i/o lag whatsoever in /mnt/A_02. These filers are 760's, by the way. Then I discovered a serious "sweet spot" at a very low number. "max" is the number of simultaneous copies between mount point pairs. "k/sec" are my visual estimates of _each_ cp rate, from watching iostat for about a minute; rates jump around by +- 100 k/sec. Column on the right is total throughput for all cp. max=3 ~ 2300 k/sec |= 6900 max=4 ~ 1800 k/sec |= 7200 max=5 ~ 1500 k/sec |= 7500 max=6 ~ 1200 k/sec |= 7200 max=8 ~ 800 k/sec |= 6400 max=9 ~ 800 k/sec |= 7200 max=10 ~ 600 k/sec |= 6000 I've seen much higher data rates from the 760's. So I'm thinking this means I need to learn nfs client tuning for solaris. Yes? Dave -- dave toal systems optimist Thomson & Thomson N. Quincy, MA -----Original Message----- From: Jeffrey Krueger To: Rainchik, Aleksandr (MED, Non GE) Cc: 'Jeffrey Krueger '; 'toasters(a)mathworks.com ' Sent: 1/15/01 1:37 PM Subject: Re: /home layout with many filers and NIS automount On Sat, Jan 13, 2001 at 03:16:20PM -0600, Rainchik, Aleksandr (MED, Non GE) wrote: > > There is another thing against > > #auto_home NIS map > > user1 filer2:/vol/vol1/& > design. We have a lot of development servers. So any time you > add/delete/modify a single user you should go trough _all_ the > NIS/automount clients and run "automount" to let automountd know > that passwd map has been just updated. Am I wrong again? This is correct, but I simple cron job which runs periodically on all NIS clients should make short work of this task. =) -- Jeff

1 0

Re: inode limitation?
by Jim VandeVegt 15 Jan '01

15 Jan '01

--- Magnus Thorne <Magnus.Thorne(a)evoice.com> wrote: > NetApp has told me that they don't recommend that I have more than > 250,000 > inodes per volume. Is this a real limitation, because my NetApps > currently > have 6.1 million inodes available? The online documentation says I > only > need to add more hard drives to get more inodes (1 per 32K). Has > anyone > used the netapp for small files? Small means millions of < 60K. > Also, I > need all these files available on one share. The training I received from NetApp personnel indicated that a particular strength of WAFL is the ability to add more inodes on the fly. There wasn't any mention of any particular recommended limit. A quick check of my system shows that even my smallest volumes currently have "Maximum Numer of Files" set over 1 million. On a 760, one volume with 330 GB of data space, 13 million on MAX files, 550 thousand currently. Another with 150 GB of data space, 6 million on MAX files, 414 thousand currently. As far as I know, no one at our site purposely expanded any of the MAX settings. ===== Jim VandeVegt vandevegt(a)yahoo.com vandevegt(a)avaya.com http://members.xoom.com/vandevegt Toilet paper should come off the front! __________________________________________________ Do You Yahoo!? Get email at your own domain with Yahoo! Mail. http://personal.mail.yahoo.com/

1 0

Re: /home layout with many filers and NIS automount
by Jim VandeVegt 15 Jan '01

15 Jan '01

--- Andrew Steingruebl <steingra(a)pprd.abbott.com> wrote: > "Rainchik, Aleksandr (MED, Non GE)" said: > > > > >Speaking about this case: > >> > >> #auto_home NIS map > >> user1 filer2:/vol/vol1/& > >> .... > >> user9999 filer3:/vol/vol4/& > > > >What I'm afraid of, is that if you have 200 users > >logged in your box, you will have 200 mountpoints, > >one mountpoint per each holy user!!! I really feel > >bad about it. Am I wrong? In the Solaris /etc/auto_home file you can specify an additional colon in the directory path to indicate the highest exported directory, and the automounter will use some sort of symbolic link such that only one connection per exported file system will result. A 'df- k', however, might still show you all the connections, I haven't tested with a large number as of yet. The syntax for /etc/auto_home for the above example would then be: user0001 filer2:/vol/vol1:& user9999 filer3:/vol/vol4:& Assuming you have /vol/vol? exported on the Filers. ===== Jim VandeVegt vandevegt(a)yahoo.com vandevegt(a)avaya.com http://members.xoom.com/vandevegt Toilet paper should come off the front! __________________________________________________ Do You Yahoo!? Get email at your own domain with Yahoo! Mail. http://personal.mail.yahoo.com/

1 0

RE: /home layout with many filers and NIS automount
by Rainchik, Aleksandr (MED, Non GE) 15 Jan '01

15 Jan '01

Speaking about this case: > > #auto_home NIS map > user1 filer2:/vol/vol1/& > .... > user9999 filer3:/vol/vol4/& What I'm afraid of, is that if you have 200 users logged in your box, you will have 200 mountpoints, one mountpoint per each holy user!!! I really feel bad about it. Am I wrong? -----Original Message----- From: Jeffrey Krueger [mailto:jkrueger@qualcomm.com] Sent: Friday, January 12, 2001 2:31 PM To: Rainchik, Aleksandr (MED, Non GE) Cc: toasters(a)mathworks.com Subject: Re: /home layout with many filers and NIS automount On Fri, Jan 12, 2001 at 10:47:54AM -0600, Rainchik, Aleksandr (MED, Non GE) wrote: > Hello, > > I just need a word of advice on how to structure /home if you > use automount, have 5-10 home directory servers (NetApp filers) > and around 1000 users. Point of clarification: 5-10 filers for 1,000 users? What model filers are they and I hope they are serving other data besides homedirs! At one point we had about 15,000 home directories mostly on two NetApps - a F330 and a F540 - and they were pretty bored filers. One way to simplify your design would be to reduce the number of filers serving homedirs, especially for 1,000 users. Two good reasons to spread out homedirs are for bandwidth constraints and to spread risk/reduce points of failure. But if you have consistently high bandwidth between all users and filers than 2-4 homedir servers should be more than plenty mitigation of risk considering the cost of additional filers versus the additional benefit (considering NetApps impressive reliability). > I'm personally is looking for /home/homeserver-vol#/user-name > scheme, so my automount maps will be nice, small and manageble: > > #auto_home NIS map > filer1-0 filer1:/vol/vol0 > filer1-1 filer1:/vol/vol1 > ... > filer5-5 filer5:/vol/vol5 This is a very restrictive design because it makes homedir moves extremely conspicuous. It also flies a bit in the face of a large selling point of the automounter - eliminating host-based network paths. If a user is moved from one filer to another, or even from one volume to another, they will have a whole new path to their home directory. Worse yet, any scripts or application configurations which use the full path to their home directory will need to be re-coded or re-configured - YUK! Being able to "hide" the location of a home directory (or any other remote filesystem) is extremely useful. > But it seems like users prefer /home/username scheme and it leads > to 1000+ NIS automount map for /home, like > > #auto_home NIS map > user1 filer2:/vol/vol1/& > .... > user9999 filer3:/vol/vol4/& This is the way we do it, and the way I'd venture to guess most/many places do it. From the user perspective, it is very simple. From the admin perspective, it is easy to manage. We've found it to give maximum flexibility with minimum hassle. > Do you know if there is more elegant solution for this problem? Can you > tell me, what is your preference for /home layout? What do you mean by elegant? The above is pretty darned elegant, and for 1,000 users it should be easily manageable too. We currently have 10,223 home directories and we've got one big home directory map like above. It works great! One recommendation is you may want to use a script or database to help generate that automount map. The easiest way would be to write a shell/perl script that scanned all your "homedir" directories on each filer, ensured username uniqueness and then build the automount map. For added flexibility, we use a small database. Each row in the homedir table has the username (mountpoint under /home), the filer, and the path to their homedir. At some interval a script comes thru and builds the map. Hope this info helps! -- Jeff -- ------------------------------------------------------------------------ ---- Jeff Krueger, NetApp CA E-Mail: jeff(a)qualcomm.com Senior Engineer Phone: 858-651-6709 NetApp Filers / UNIX Infrastructure Fax: 858-651-6627 QUALCOMM, Inc. IT Engineering Web: www.qualcomm.com

3 2

Can volumes span FCAL controllers?
by Matt Phelps 13 Jan '01

13 Jan '01

Hi, We're probably going to ba adding another FCAL controller to our 840 soon. Can volumes include disks that are on both controllers? -- Matt Phelps System Administrator, Computation Facility Harvard - Smithsonian Center for Astrophysics mphelps(a)cfa.harvard.edu, http://cfa-www.harvard.edu/~mphelps

6 9

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