toasters

toasters@lists.teaparty.net

2 participants
13529 discussions

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

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