toasters

toasters@lists.teaparty.net

13515 discussions

AIX over FC speed issues
by Page, Jeremy 07 Oct '08

07 Oct '08

We've replaced our DS4500 and EMC systems with an IBM rebranded 3070 cluster. For the most part I'm very pleased with the performance but for a specific application the new system is about 15% slower then the old DS4500 was, even though it's got more spindles at the same speed then originally. Old system (DS4500) (15k FC) 8 disks in a RAID 5 array 10 in a RAID 10 array 2 more mirrored New system (3070A) (15k FC) < This box is idling, the only stuff on it is this application (most stuff is on the other controller since I knew this app was a pain) 3 12 disk RAID groups in a single aggr (DP) Should be 20 versus 30 spindles. I understand that there may be a slight impact from DP versus RAID0/10 for parity calculations, but I expected it to be more then absorbed by the 16x size cache on the Netapp, the extra spindles and the NVRAM for writes. So any suggestions on how to speed this up would be useful. I'm going to have the AIX guy re-run his tests tonight while I have statit running to capture how the disks are impacted but I think possibly the bottleneck is elsewhere. Below are the test results from Sunday. IBM SAN Stats (READ) Benchmark Statistics for Database DB Name: /vrback1/vrdata1/veeder Proc Date: 10/05/08 Reading part master file Records Start End Duration ---------- -------- -------- -------- pt_mstr 52,674 12:50:30 12:50:52 00:00:22 reading customer master file Records Start End Duration ---------- -------- -------- -------- cm_mstr 17,984 12:50:52 12:51:14 00:00:22 reading order headers Records Start End Duration ---------- -------- -------- -------- so_mstr 4,213 12:51:20 12:51:28 00:00:08 reading order detail Records Start End Duration ---------- -------- -------- -------- sod_det 11,210 12:51:28 12:51:41 00:00:13 reading routing records Records Start End Duration ---------- -------- -------- -------- ro_det 87,514 12:51:52 12:52:13 00:00:21 reading address master file Records Start End Duration ---------- -------- -------- -------- ad_mstr 279,824 12:52:13 12:57:14 00:05:01 read install base records Records Start End Duration ---------- -------- -------- -------- isb_mstr 639,999 12:57:19 13:06:18 00:08:59 Read invoice detail file Records Start End Duration ---------- -------- -------- -------- idh_hist 4,573,493 13:06:18 13:57:25 00:51:07 NetAPPSAN Stats (READ) Benchmark Statistics for Database DB Name: /vrdata1_mr/veeder Proc Date: 05/10/08 Reading pt_mstr Records Start End Duration ---------- -------- -------- -------- pt_mstr 52,674 12:43:08 12:43:43 00:00:35 reading customer master file Records Start End Duration ---------- -------- -------- -------- cm_mstr 17,984 12:43:43 12:44:12 00:00:29 reading order headers Records Start End Duration ---------- -------- -------- -------- so_mstr 4,213 12:44:19 12:44:35 00:00:16 reading order detail Records Start End Duration ---------- -------- -------- -------- sod_det 11,210 12:44:35 12:44:57 00:00:22 reading routing records Records Start End Duration ---------- -------- -------- -------- ro_det 87,514 12:45:00 12:45:27 00:00:27 reading address master file Records Start End Duration ---------- -------- -------- -------- ad_mstr 279,824 12:45:27 12:51:48 00:06:21 read install base records Records Start End Duration ---------- -------- -------- -------- isb_mstr 639,999 12:51:55 13:02:56 00:11:01 Read invoice detail file Records Start End Duration ---------- -------- -------- -------- idh_hist 4,573,493 13:02:56 13:47:13 00:44:17 IBM SAN Stats (Write) R,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,T .wrote 100,000 records . ./spare3/qad86edb/bmark . . Start: 09:04:09 . . End: 09:05:30 . . Duration: 00:01:21 . F,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,G NetAPPSAN Stats (Write) R,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,T .wrote 100,000 records . ./spare5_mr/bmark . . Start: 09:09:38 . . End: 09:11:14 . . Duration: 00:01:36 . F,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,G This message (including any attachments) contains confidential and/or proprietary information intended only for the addressee. Any unauthorized disclosure, copying, distribution or reliance on the contents of this information is strictly prohibited and may constitute a violation of law. If you are not the intended recipient, please notify the sender immediately by responding to this e-mail, and delete the message from your system. If you have any questions about this e-mail please notify the sender immediately.

3 4

Re: Synchronous SnapMirror destinations + NDMP moves
by Steve Rieger 07 Oct '08

07 Oct '08

Yup you are right ------Original Message------ From: George T Chen Sender: owner-toasters(a)mathworks.com To: Nick Silkey To: toasters(a)mathworks.com Sent: Oct 6, 2008 15:13 Subject: RE: Synchronous SnapMirror destinations + NDMP moves I would have to try it out myself, but I'm under the impression that a synchronous snapmirror destination cannot be the source of a next cascade (because it can't create a snapshot). > -----Original Message----- > From: owner-toasters(a)mathworks.com [mailto:owner-toasters@mathworks.com] > On Behalf Of Nick Silkey > Sent: Monday, October 06, 2008 1:27 PM > To: toasters(a)mathworks.com > Subject: Synchronous SnapMirror destinations + NDMP moves > > (NB: Resent from $oldjob. Newly-subscribed identity is having issues > posting.) > > Toasters -- > > Anyone out there NDMP moving data from synchronous SnapMirror > destinations successfully? We have a FAS3020C cluster at 7.2.4 and > are moving forward on each head synchronously SnapMirroring via GigE > to a remote FAS3020, also at 7.2.4. Ultimately we want to cut backups > via NDMP to TSM over FCP from the SnapMirror destinations on secondary > storage. There are some concerns which were raised by our SE about > SnapMirror and NDMP stepping on one another, > <http://now.netapp.com/Knowledgebase/solutionarea.asp?id=kb44293> > being one example. > > Looking to hear from those in the trenches on the above. Thanks. > > -- > Nick Sent via BlackBerry from T-Mobile

1 0

PRAGUE CONFERENCE
by Milazzo Giacomo 06 Oct '08

06 Oct '08

Hi all Who of you will be presente at NetApp "guru" conference in Prague next november 17-19? My boss sponsorized it for me and a collegue of mine. J

1 0

DECRU FTP ACCESS BUG?
by Milazzo Giacomo 06 Oct '08

06 Oct '08

Hi all A customer of mine asked me to set up ftp access on his E-Series cluster already acting as NFS NAS. OS is 3.1.1.1 I've followed instructions and it seems to work but there's a strange thing and something not clear. Once created the volume, exported, created virtual share, added the cryptainer I've enabled the ftp this way: >vip set -ftp-access on vnetapp1 In the guy I've created a user on the share called ftplogger with password and then I've set up his home directory this way >user home set ftplogger@app_domain rmrcrsdec-01:/vol/ciscolog (the domain has internal users only, no NIS or other external sources) Last I've enabled two hosts for test purposes, the first enabling both read and write and the other one only read (using the IP access functionality) What I've noted is that I can log on Decru using any username and password!!! My rights depend only from the IP. So that I can open a r/w ftp connection from the enabled IP address also using johndoe as user and any password!!! (I got this because once I was sure to wrong to input the password and I expected some 'access denied'...instead I log in!) FTP home dir seems to work fine, once logged user is pointed in the right share and if he tried to change he got access denied but if I perform the command: >user home list I can always read root@app_domain as user using rmrcrsdec-01:/vol/ciscolog and not the ftplogger user created. Thank you in advance Dott. Giacomo Milazzo Technical Account Manager Sinergy SpA Filiale di Roma + 00148. viale Castello della Magliana, 38 ' (+39) 3406001045 0665970252 7 +39 02 26922048 - Giacomo.Milazzo(a)Sinergy.it

1 0

san thin provisioning best practices doc
by Nicholas Bernstein 02 Oct '08

02 Oct '08

I'm looking for a good best practices doc I can point students to as a takeaway on using a fractional reserve with luns. Could anyone point me to a good doc? I looked on the now site and the best thing I came across was tr3483, but that just talks about how the process works as opposed to taking about best practices. Thanks, Nick -- Nicholas Bernstein Technologist, Consultant, Instructor http://nicholasbernstein.com

1 0

Operations Manager Options on FAS devices
by Jimmy Corncrake 02 Oct '08

02 Oct '08

Greetings all, Though FilerView we can set a number of Operations Manager options under the 'Operations Manager[image: Right]Access' tabs. These options include, *Operations Manager Links: **DataFabric Manager server:* *DataFabric Manager Protocol:* *DataFabric Manager Port:* ** I would like to set these options on the CLI but after looking and looking I can not find which command \option etc contains these. Anyone know? Perhaps it something I need to push from Operations Manager but this would surprise me somewhat. rgds, Jimmy

1 0

OnTap Simulator
by Rob Borowicz 01 Oct '08

01 Oct '08

I'm doing some development against SNMP in DOT and am using the just posted Sim for 7.3. Anybody care to privately exchange notes on their Simulator experiences? TIA -Rob

3 2

RE: issues with 10G Ethernet.
by Learmonth, Peter 30 Sep '08

30 Sep '08

Sure thing, Jeremy. I talked to NetApp engineering (the networking group). I was doing my homework prepping for my VMWorld breakout session,ran across that note and said "That can't be right"...so I checked with engineering. They said the docs wil be fixed. Share and enjoy! Peter -----Original Message----- From: Page, Jeremy <jeremy.page(a)gilbarco.com> Sent: Tuesday, September 30, 2008 9:27 AM To: Timothy Hollingworth <thollingsworth(a)EPLUS.com>; toasters(a)mathworks.com <toasters(a)mathworks.com> Subject: RE: issues with 10G Ethernet. I'd expect lower latency doing the TCP stuff with the cards ASIC then having to wait for a general purpose processor to take care of it. I am using Etherchannel (instead of LACP), I was also under the impression that LACP is not supported with 10gE VIFs, Peter can you tell us where you heard otherwise? Jeremy M. Page____________________ Systems Architect * email:Jeremy.Page@gilbarco.com - * phone: 336.547.5399 - 6 fax: 336.547.5163 - * cell: 336.601.7274 ________________________________ From: owner-toasters(a)mathworks.com [mailto:owner-toasters@mathworks.com] On Behalf Of Timothy Hollingworth Sent: Monday, September 29, 2008 6:22 PM To: toasters(a)mathworks.com Subject: RE: issues with 10G Ethernet. Page 121 of the Network Management Guide for DataONTAP 7.3 states... The 10GbE TOE NIC cards have a number of limitations when used with vifs. They include: * Multimode vif limited to two (2) 10GbE TOE NICs * LACP not supported with 10GbE TOE NICs * TOE functionality disabled on 10GbE NIC in vif In a multimode vif, DataONTAP controls only outbound packets. The link used for inbound packets is decided upon by the device at the other end of the vif. I'm not an expert in this area but assume that TOE won't work because each TOE card wouldn't know what packets the other has processed in terms of inbound traffic. If the processor on the filer is not being maxed out, wouldn't it make sense to turn off TOE anyway and just let the filer do the processing? -Tim Hollingworth- -ePlus Technology Inc.- From: owner-toasters(a)mathworks.com [mailto:owner-toasters@mathworks.com] On Behalf Of James Beal Sent: Monday, September 29, 2008 12:16 PM To: Ed Wilts Cc: Page, Jeremy; toasters(a)mathworks.com Subject: Re: issues with 10G Ethernet. On 29 Sep 2008, at 16:36, Ed Wilts wrote: >From what I've heard, you can use TOE but only in an active/passive configuration. So if you have 2 10GigE ports on your card with each going to a separate switch, as long as you only use 1 active path at a time, you'll be fine. You'll have full failover and TOE, but you won't have 20GigE of bandwidth available. On Mon, Sep 29, 2008 at 9:34 AM, Page, Jeremy <jeremy.page(a)gilbarco.com> wrote: What an unfortunate fix :( I'd very much like to use TOE on my 10gig VIF, does anyone know if this is ever going to be supported, or is there an architectual reason why it won't work? Just for clarification if I have TOE turned on with 10G ethernet my units stop serving data with no cluster failover occurring. At least with 7.2.5.1 with TOE turned off the units appear to be stable previously they where unstable whether TOE was active or not. This message (including any attachments) contains confidential and/or proprietary information intended only for the addressee. Any unauthorized disclosure, copying, distribution or reliance on the contents of this information is strictly prohibited and may constitute a violation of law. If you are not the intended recipient, please notify the sender immediately by responding to this e-mail, and delete the message from your system. If you have any questions about this e-mail please notify the sender immediately.

1 0

Re: issues with 10G Ethernet.
by James Beal 30 Sep '08

30 Sep '08

On 29 Sep 2008, at 16:36, Ed Wilts wrote: > From what I've heard, you can use TOE but only in an active/passive > configuration. So if you have 2 10GigE ports on your card with each > going to a separate switch, as long as you only use 1 active path at > a time, you'll be fine. You'll have full failover and TOE, but you > won't have 20GigE of bandwidth available. > > On Mon, Sep 29, 2008 at 9:34 AM, Page, Jeremy <jeremy.page(a)gilbarco.com > > wrote: > What an unfortunate fix :( > > I'd very much like to use TOE on my 10gig VIF, does anyone know if > this > is ever going to be supported, or is there an architectual reason > why it > won't work? > Just for clarification if I have TOE turned on with 10G ethernet my units stop serving data with no cluster failover occurring. At least with 7.2.5.1 with TOE turned off the units appear to be stable previously they where unstable whether TOE was active or not.

3 2

RE: issues with 10G Ethernet.
by Learmonth, Peter 30 Sep '08

30 Sep '08

Hi I had previously checked on this and the second bullet is incorrect. LACP _is_ supported with 10GbE TOE cards. Share and enjoy! Peter -----Original Message----- From: Timothy Hollingworth <thollingsworth(a)EPLUS.com> Sent: Monday, September 29, 2008 3:44 PM To: toasters(a)mathworks.com <toasters(a)mathworks.com> Subject: RE: issues with 10G Ethernet. Page 121 of the Network Management Guide for DataONTAP 7.3 states... The 10GbE TOE NIC cards have a number of limitations when used with vifs. They include: * Multimode vif limited to two (2) 10GbE TOE NICs * LACP not supported with 10GbE TOE NICs * TOE functionality disabled on 10GbE NIC in vif In a multimode vif, DataONTAP controls only outbound packets. The link used for inbound packets is decided upon by the device at the other end of the vif. I'm not an expert in this area but assume that TOE won't work because each TOE card wouldn't know what packets the other has processed in terms of inbound traffic. If the processor on the filer is not being maxed out, wouldn't it make sense to turn off TOE anyway and just let the filer do the processing? -Tim Hollingworth- -ePlus Technology Inc.- From: owner-toasters(a)mathworks.com [mailto:owner-toasters@mathworks.com] On Behalf Of James Beal Sent: Monday, September 29, 2008 12:16 PM To: Ed Wilts Cc: Page, Jeremy; toasters(a)mathworks.com Subject: Re: issues with 10G Ethernet. On 29 Sep 2008, at 16:36, Ed Wilts wrote: >From what I've heard, you can use TOE but only in an active/passive configuration. So if you have 2 10GigE ports on your card with each going to a separate switch, as long as you only use 1 active path at a time, you'll be fine. You'll have full failover and TOE, but you won't have 20GigE of bandwidth available. On Mon, Sep 29, 2008 at 9:34 AM, Page, Jeremy <jeremy.page(a)gilbarco.com<mailto:jeremy.page@gilbarco.com>> wrote: What an unfortunate fix :( I'd very much like to use TOE on my 10gig VIF, does anyone know if this is ever going to be supported, or is there an architectual reason why it won't work? Just for clarification if I have TOE turned on with 10G ethernet my units stop serving data with no cluster failover occurring. At least with 7.2.5.1 with TOE turned off the units appear to be stable previously they where unstable whether TOE was active or not.

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