toasters August 2001

toasters@lists.teaparty.net

122 participants
161 discussions

Oracle performance on filers
by John Hope-Bailie 30 Aug '01

30 Aug '01

Greetings, We recently started testing an F740 in a Sun Solaris 2.6 environment with Oracle 8 The host is a Sun E4000 and has a Sun A1000 disk attached. Normally, the filers perform far better than direct-attached disk in this setup. However when we run all the Oracle stuff on the filer, including the directory specified by ORACLE_HOME, the system is sorrowfully slow. We have another login, which points ORACLE_HOME at a location on the A1000 storage. The Oracle datafiles are all still located on the filer. In this case, the system is normal (i.e. good and fast). I am at a loss to explain why the location of the ORACLE_HOME would affect things in this way. During the poor Oracle performance, activity on the filer is minimal (3 - 5%) When we throw bursts of large file data at the filer (copy from /dev/zero) we get throughputs of about 12 MB/s at utilisations of 50 to 80 percent so the filer seems to be working fine. Can anyone suggest where to look next ? regards John Hope-Bailie source consulting E-mail: johnhb(a)sourceconsulting.co.za

1 0

RE: access filer from boot floppy [getting OT]
by Parsons Alastair 30 Aug '01

30 Aug '01

All, I am not sure whether this is relevant to your problem but if you are using Norton Ghost to send or receive images from the Filer there is an issue with TCP Frame Size which the Filer and Norton Ghost use. We have had to set up a Cifs Alias to our Filer with a different Frame Size specifically to Solve the 'Ghost being Slow' issue. The details of the bug and its fix are in Bug Report 27586 on the NOW website. Cheers Alastair -----Original Message----- From: Mumphrey, Robert N [mailto:Robert.Mumphrey@AIMFUNDS.COM] Sent: 29 August 01 18:26 To: 'Witham, John'; Mumphrey, Robert N; 'Asbjørn Bjørkelund'; 'toasters(a)mathworks.com' Subject: RE: access filer from boot floppy [getting OT] We send our images directly to the filer. One problem we initially had was that you need to make sure that the NIC card speed and duplex on the PC is not set for Auto Detect. Set the Speed and Duplex manually in the protocol.ini on the floppy. Each NIC is different in how you need to set this, so you will need to find out what works for your NIC. Also, on the ghost command line, we needed to specify a different buffer size than the default. We use a buffer size of 8 and it seems to work fine. If we left it default, it was extremely slow. Have a great day, Robert N. Mumphrey Associate Principal Information Technology Services AIM Funds Phone: (800) 347-1919 ext. 7137 Fax: (713) 986-9732 E-mail: Robert.Mumphrey(a)Aimfunds.com -----Original Message----- From: Witham, John [mailto:jwitham@takedapharm.com] Sent: Wednesday, August 29, 2001 11:55 AM To: 'Mumphrey, Robert N'; 'Asbjørn Bjørkelund'; 'toasters(a)mathworks.com' Subject: RE: access filer from boot floppy [getting OT] There's a flawless boot disk at http://www.bovistech.com/detect.htm <http://www.bovistech.com/detect.htm> . It will autodetect most network cards. We use it for GHOSTing our client PCs. On a similar note, we have severe problems _sending_ and image to the filers from any NetBIOS boot disk. I've read that it's DOS limitations to NetBIOS, but am not sure. Presently, I use a NetBeui boot disk to send the image to an NT workstation and then copy it to the filer... Works, but is a kludge. A nybody doing anything different? -john John Witham Senior Data Networking Engineer Takeda Pharmaceuticals North America, Inc. V://847.383.3304 F://847.383.3205 mailto://jwitham@takedapharm.com <mailto://jwitham@takedapharm.com> -----Original Message----- From: Mumphrey, Robert N [ mailto:Robert.Mumphrey@AIMFUNDS.COM <mailto:Robert.Mumphrey@AIMFUNDS.COM> ] Sent: Wednesday, August 29, 2001 9:04 AM To: 'Asbjørn Bjørkelund'; 'toasters(a)mathworks.com' Subject: RE: access filer from boot floppy We are performing this exact process at our site. We actually do our Windows 2000 automated server loads from our filer using a boot floppy. Are you using a standard type of boot floppy created using Microsoft's NetworkD isk Creation utility that gets loaded with NT Server 4.0? As long as yous pecify the WINS servers (and the filers have WINS entries) in the protocol.ini, you should be able to map drives straight away. Are you ablet o get on the network at all with your floppy? Are you authenticating tot he domain? Hope this helps some. -----Original Message----- From: Asbjørn Bjørkelund [ mailto:asbjorn.bjorkelund@proact.no <mailto:asbjorn.bjorkelund@proact.no> ] Sent: Wednesday, August 29, 2001 3:52 AM To: 'toasters(a)mathworks.com' Subject: access filer from boot floppy Hello guru's We have several workstations which we want to take ghost image of, and store it on the filer, but when i try to access the filer from a boot floppy I'am not able to access the filer. I have tried to map network drive to a unix and ntfs style volume but nothing helps. I have changed the user rights for the volume, but it did not help. I canm ap network drives to other nt servers we have, but not to the filer. Is it possible to do this on a filer? I hope somebody can help me out here. Regards Asbjorn Bjorkelund ProAct Systems Phone + 47 51601460 Mobile + 47 932 30623 This message is for the designated recipient only and may contain privileged or confidential information. If you have received it in error, please notify the sender immediately and delete the original. Any other use of the email by you is prohibited. ********************************************************************** This email and any attachments may be confidential and the subject of legal professional privilege. Any disclosure, use, storage or copying of this email without the consent of the sender is strictly prohibited. Please notify the sender immediately if you are not the intended recipient and then delete the email from your inbox and do not disclose the contents to another person, use, copy or store the information in any medium. **********************************************************************

1 0

RE: Oracle hot backups and snapshots
by Yarmas, Tom 29 Aug '01

29 Aug '01

> 1) The TR defines a sequence in which all tablespaces are > placed into hot backup mode. It has been suggested that this step might take > an extended period of time. Our tests to date have been on relatively small data > volumes and the "ALTER TABLESPACE <name> BEGIN BACKUP;" steps > have been almost instantaneous. > Is it likely that this step would end up taking longer (say > tens of minutes) for large DBs ? > No, the Alter Tablespace command will always be virtually instantaneous. It does not depend on the size of the tablespace. That command is simply setting a "flag" in the control file and in the database so that the database is aware of what datafiles are being backed up, and so additional logging occurs for transactions associated with those datafiles. Of course if you have hundreds of tablespaces, then the sum can add up. In other words, hundreds of tablespaces could take more than a couple of seconds. The good news is that with ONTap version 6.x, the SnapShot creation is virtually instantaneous as well, so even if it took 60 seconds to alter all the tablespaces, the database would be in hot backup mode only a minute or so. > During testing we have found that recovering the entire > snapshot (with all data files from all schemas) works fine. But when we have > recovered only the single datafile related to one schema, we have run into > problems. I feel we may be overlooking some basic issue here. Not sure what you might be overlooking. Oracle is extremely recoverable. In order to accomplish what you want, it should be very straightforward. I have recovered individual datafiles and tablespaces with SnapShots many times with no problems. Doing that should not affect any other users, providing that individual schemas are in seperate tablespaces. You need to take the affected tablespace offline, replace the affected datafile(s) from the proper SnapShot directory (simply copy them over the originals) then run recovery using either the "datafile" or "tablespace" directive. As in alter database recover datafile '/oradata/foo.dbf'; or alter database recover tablespace foo; You can issue either of those commands when the database is open providing that the tablespace is offline. Hope that helps. -tom

1 0

NFS read performance
by kevin graham 29 Aug '01

29 Aug '01

I'm caught in a loop of overthinking things I don't know in enough depth... Under Solaris, are the filesystem drivers smart enough to control EWOULDBLOCK based on nfs read-ahead's? I'm trying to decide if it'd be worth going to the trouble of implementing nonblocking io into the file reads of our httpd (whose docuement tree is all nfs-mounted)... Given the inherent latencies in NFS vs. local disk, I'm hoping that I could only issue reads to filehandles that already have read-ahead data. There's definately alot of other stuff that has to be worked and thought out as well, but I thought I figured I first see if the basic functionality was there before digging into it harder.. thanks. ..kg..

5 4

RE: Oracle hot backups and snapshots
by John Hope-Bailie 29 Aug '01

29 Aug '01

Thanks for the response. See below -----Original Message----- From: Penelope Carr [mailto:Penelope.Carr@veritas.com] Sent: 29 August 2001 09:09 To: John Hope-Bailie Subject: RE: Oracle hot backups and snapshots When you say run into problems retrieving one datafile what do you mean? What tool are you using to do the hot backup? This will determine if you can do what you are asking. We do the hot backup as described in the NetApp TR 3049, i.e. run a script on the Sun host which does the following ;- 1) Issues an "alter tablespace <name> begin backup" command via sysmgrl for every tablespace (puts them into hot backup mode) 2) Deletes the oldest snapshot on the filer and bumps the others along by 1 (simply renames them) 3) Takes a new snapshot on the filer 4) Issues an "alter tablespace <name> end backup" command via sysmgrl for every tablespace (takes them out of hot backup mode) We have run a few simple tests thus far for restore of single tablespaces in which we :- 1) Truncated a tablespace to effectively "break" it 2) Took the tablespace offline 3) Recovered the previous version of the relevant data file from the snapshot (by a simple copy) 4) Brought the tablespace online ( All this time the DB and other tablespaces were up) The tablespace requires media recovery at this point. Once this command is run, it seems to be o.k. with all rows present. I am unsure about what is hapening here. An explanation would be welcome. When tablespaces are put into backup mode they are all done at once and should not take a huge amount of time to do so. It is the backup that takes longer as the database grows. There is tuning in the database that can be done to help the speed of the backup. Penelope Penelope Carr cell 914-439-4911 office 914-934-2404 -----Original Message----- From: John Hope-Bailie [mailto:johnhb@global.co.za] Sent: Wednesday, August 29, 2001 2:19 PM To: toasters(a)mathworks.com Subject: Oracle hot backups and snapshots Greetings, We are trying to utilise a NetApp filer's snapshot feature to enable rapid recovery in an Oracle DB environment. We are aware of the www.netapp.com/tech_library/3049TR document which covers the basics. The intention is to use the hot backup alternative to create several snapshots throughout the day to provide maximum coverage. Both Oracle 8 and 8i are in use. I have 2 questions :- 1) The TR defines a sequence in which all tablespaces are placed into hot backup mode. It has been suggested that this step might take an extended period of time. Our tests to date have been on relatively small data volumes and the "ALTER TABLESPACE <name> BEGIN BACKUP;" steps have been almost instantaneous. Is it likely that this step would end up taking longer (say tens of minutes) for large DBs ? 2) The system has been set up so that each "customer data set" is defined as a separate schema, many of which are contained within a single database. Each schema contains one or more data files, but these only contain data related to their own schema and not other schemas. The idea here is that if it was necessary to recover data related to one customer, it should be possible to do this without taking any other customer tablespaces offline. During testing we have found that recovering the entire snapshot (with all data files from all schemas) works fine. But when we have recovered only the single datafile related to one schema, we have run into problems. I feel we may be overlooking some basic issue here. Any advice would be welcome. regards, John Hope-Bailie Source Consulting, South Africa email: johnhb(a)sourceconsulting.co.za

1 0

RE: NDMP for Solaris
by Ken Hillier 29 Aug '01

29 Aug '01

NDMP can not, at this time, be used to connect to a client other than a filer. The dump/restore mechanism is written to interface with the NMDPD on the filers. This functionality would greatly increase NDMPCopy's usefulness. You mention rsync with an NFS transfer. Have you looked into rdist as well? I have used that utility on Sun with great success. Eitherway, it souds like you have a viable plan in place. HTH I need to migrate all the date from a T3 attached to SUN box to a NetApp filer. I know ndmpcopy is a great fast tool. Anybody knows way to use it for a sun-to-netapp migration..? Any other alternatives welcome.. Needs to do this with minimum downtime. I plan to do a NFS transfer and then rsync, but looking for a faster-smarter way. TIA __________________________________________________ Do You Yahoo!? Get email alerts & NEW webcam video instant messaging with Yahoo! Messenger http://im.yahoo.com

1 0

Non-root /etc directories
by Jeffrey Krueger 29 Aug '01

29 Aug '01

Can anyone help me identify what can't be deleted from the /etc of a non-root volume? I know that all volumes get some info stuffed in there sooner or later. We just snapmirrored a few root volumes to non-root volumes and would like to get rid of the unnecessary /etc files. Mainly, it looks like it uses /etc/oldmaps and /etc/tmp for backups or something. -- Jeff

1 0

NDMP for Solaris
by Premanshu Jain 29 Aug '01

29 Aug '01

I need to migrate all the date from a T3 attached to SUN box to a NetApp filer. I know ndmpcopy is a great fast tool. Anybody knows way to use it for a sun-to-netapp migration..? Any other alternatives welcome.. Needs to do this with minimum downtime. I plan to do a NFS transfer and then rsync, but looking for a faster-smarter way. TIA

1 0

RE: Sample outputs? (Please help!)
by l l 29 Aug '01

29 Aug '01

Sorry, but I am having email problems. Please send those sample outputs to my alternate email harryxu41(a)hotmail.com. Again, I really appreciate your help! -- _______________________________________________ FREE Personalized E-mail at Mail.com http://www.mail.com/?sr=signup Talk More, Pay Less with Net2Phone Direct(R), up to 1500 minutes free! http://www.net2phone.com/cgi-bin/link.cgi?143

1 0

Re: Clustering question.
by ScottUCE＠aol.com 29 Aug '01

29 Aug '01

In a message dated 8/28/01 4:36:08 PM Central Daylight Time, rusty.dehorn(a)netapp.com writes: << This is no problem. Contact your Inside Sales rep for the necessary hardware and software additions to make it happen. If you don't know who to call, let me know. Rusty 408/822-6859 >> FYI, we have a clustered pair of F760s available with transferable NFS, CIFS and Cluster licenses. We can configure the hardware to suit. Scott Fischmann Union Computer Exchange, Inc. 6233 Idylwood Lane Minneapolis, MN 55436 Phone: 952 935 7282 FAX: 952 935 5056 scott(a)unioncomputer.com

1 0

Jump to page:

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 August 2001