toasters

toasters@lists.teaparty.net

3 participants
13514 discussions

Re: directory optimisation.
by jrg＠gbnet.net 08 Jun '99

08 Jun '99

On Tue 8 Jun, 1999, yann.golanski(a)ThePLAnet.net wrote: >/vol/vol1/data/Z/XX/XXXX/<full_usr_name>/files which reminds me - is there any concept akin to 'toolong' in the filer name lookup cache? You probably need to watch out for that at your nfs client too. >where XX and XXXX are the first 2 and first 4 characters of > <full_usr_name> and > Z is a hash spliting the <full_usr_name> between 2 filers -- yes, we > have a rather large system ;) why not just use the hash the username, and then split that up into a suitable number of chunks? James.

1 0

Q: lose shelf power supply
by George Kahler 08 Jun '99

08 Jun '99

Could someone please explain what happens when you loose a single power supply in a shelf? Is the filer smart (dumb) enough to halt? Do I get 'double disk error' and loose the filesystem? There is a discussion going on here as to what happens in a hardware RAID configuration vs a NetApp filer when a power supply fails. The thought is that if I have only one power supply in a shelf and it fails (failing 7 disks) it is like having multiple disks fail and therefore you loose the filesystem. Thanks, George ------------------------------------------------------------------------------- George Kahler e-mail: george(a)yorku.ca UNIX Systems Administrator humans: (416) 736-5257 x.22699 Computing Services, York University machines: (416) 736-5830 Ontario, Canada, M3J-1P3

5 4

Re: directory optimisation.
by Jim McCoy 08 Jun '99

08 Jun '99

>Suppose I have a filer (760) with about 50 million files -- between >1k and 500M. What would be the optimum directory structure to keep them all? We had the same problem at Yahoo when it comes to storing user database entries (e.g. MyYahoo account info) and mailboxes for YahooMail accounts. The biggest win we found was that you should go to great lengths to make sure that your directories can stay in the cache as much as possible. For YahooMail this was a four layer structure (/##/##/####/<hash>/maildir) and for the MyYahoo it was three layers. The WAFL system is pretty good when it comes to doing lookups on big directories, the only thing you need to do to help it out is to load up on system RAM and keep directories small enough that you can keep lots of them in the filers cache so that with a little bit of luck lookups run through the directories in cache and then hit the disk for the actual files. jim mccoy Yahoo! Inc.

1 0

Re: DLT library controls
by Jay Sekora 08 Jun '99

08 Jun '99

> That brings up a related question though -- since ndmp allows SCSI > commands to be sent directly to the device, that would theoretically allow > one to control the robotics over the network via NDMP. > > Do any of the NDMP-based packages support this? (Legato, Budtool, > whatever) BudTool (running, obviously, on a Unix box) does indeed tell our Breece Hill tape library (attached to the NetApp) to load and unload specific tapes, and it talks to the library through the NetApp via NDMP. So I guess this must be how it's doing it. Cheers, Jay Sekora <jay(a)ccs.neu.edu> Unix Systems Administrator Northeastern University College of Computer Science

1 0

DLT library controls
by cng1＠cecil.muscat.com 08 Jun '99

08 Jun '99

I've had a DLT7000 tape library for some time now backing up our ageing 540 but I've never been terribly happy with the controls for loading/unloading the library. In order to backup start a backup on a given tape I have to either go to the loaders console and tell it to load the tape or I need to know what tape I am on and give the correct number of "mt -f urst0a status" commands to cycle through to the correct tape. I believe that I could use something other than dump to do the backups but dump is free and does exactly what I want. The problem is compounded that since upgrading the firmware on the tape drive (to allow autocleaning) the mt -f urst0a status trick hasn't worked. Is there a better way? I would've thought that there should be some better drive robotics control functions - have they crept in between 4.3 and 5.x? FYI I'm running 4.3 connected to a overland dlt7100 drive. Chris -- Chris Good - Muscat Ltd. The Westbrook Centre, Milton Rd, Cambridge UK Phone: 01223 715006 Mobile: 07801 788997 http://www.muscat.com

4 4

RE: DLT library controls
by Paul English 08 Jun '99

08 Jun '99

> I believe that you can set the Overland Data library to run in sequential > "stacker" mode. > This mode will automatically change tapes when they are full. > There are no robotics controls in DataONTAP. > That brings up a related question though -- since ndmp allows SCSI commands to be sent directly to the device, that would theoretically allow one to control the robotics over the network via NDMP. Do any of the NDMP-based packages support this? (Legato, Budtool, whatever) Has anyone gotten this working by hacking around with NDMP? -- Paul English (Tallpaul) (508) 647-7715 tallpaul(a)mathworks.com (508) 647-7015 FAX Walkie Channel 2 ----- End Included Message -----

1 0

directory optimisation.
by yann.golanski＠ThePLAnet.net 08 Jun '99

08 Jun '99

Suppose I have a filer (760) with about 50 million files -- between 1k and 500M. What would be the optimum directory structure to keep them all? So far I have the following: /vol/vol1/data/Z/XX/XXXX/<full_usr_name>/files where XX and XXXX are the first 2 and first 4 characters of <full_usr_name> and Z is a hash spliting the <full_usr_name> between 2 filers -- yes, we have a rather large system ;) This seems to be working fairly well, but I am wondering if there isn't a better way of doing it. Does anyone knows how Netcaches keep their URLs in memory? Could you use that methode for the above? Any ideas? -- Yann Golanski Internet Systems Developer yann.golanski(a)theplanet.net The Planet Online

1 0

RE: Metadata preservation during network file backup
by Benn, Paul 08 Jun '99

08 Jun '99

NT metadata is lost when backing up via NFS. If you perform a backup via CIFS of files with UNIX-style permissions, the permissions are saved and restored. Additionally, WAFL keeps a DOS-like archive bit on UNIX files (as well as NT files). NT based backup apps can leverage this to perform incremental backups of UNIX files. You can check out my document on NT backup at: http://www.netapp.com/technology/level3/3052.html Regards, Paul -----Original Message----- From: Itzik (Itzhak) Meirson DSc. [mailto:imeirson@xpert.com] Sent: Tuesday, June 08, 1999 12:20 AM To: toasters(a)mathworks.com Subject: Metadata preservation during network file backup I would appreciate if someone could clarify the issue of metadata preservation during network backup of filers data. Lets assume that I have a share with NFS/CIFS access. If backing up this share from NFS will the CIFS metadata preserved upon restore? If backing up this share from CIFS will the NFS ACLs be preserved upon restore? Thnaks, Itzik

1 0

Re: Q: lose shelf power supply
by Jim McCoy 07 Jun '99

07 Jun '99

Michael Alvarado writes: >One other point to emphasize is that power supplies are fairly reliable devices >and a redundant supply can be purchased for even greater protection. If fact, in my experience the power supplies are far more reliable than the shelves themselves. Having gone through such events before (shelf failures, PS failures, etc) I can assure you that the system does "the right thing" and once you have replaced the failed hardware everything starts running again just fine. jim mccoy Yahoo! Inc.

1 0

Filer Impressions, please.
by Nathan O. Siemers 07 Jun '99

07 Jun '99

Hello, I have joined this list because I am considering a purchase of either (~ 1/2 tB) : 1. A Netapp 760 Filer. 2. An Auspex file server. 3. Beefed up local RAID. to serve as common, high performance, *shared* file space for two Origin2000 servers. I imagine a dedicated high speed network between the toaster and the two origins will be put in place if we go with the netapp type solution. I would appreciate your frank comments on the Filer's performance and reliability. Thanks, nathan -- N a t h a n O . S i e m e r s Bioinformatics Division of Applied Genomics Bristol-Myers Squibb Pharmaceutical Research Institute Hopewell Building 3B, P.O. Box 5400, Princeton, NJ 08543-5400 609 818-6568 siemersn(a)bms.com

4 3

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