toasters

toasters@lists.teaparty.net

13513 discussions

Another SSD question
by Basil 07 Jul '15

07 Jul '15

So I'm looking at upgrading a few 8020s and trying to choose between flashpool and flashcache. Flashpool is more expensive, even with just 12x200 SSD drives, and we'd need to find space in a very tight rack for it. The workload is 80% exchange and 20% VMWare, and we would like to use deduplication without snapshots. Is it possible to use PCS to compare flashcache and flashpool benefits? How can I translate the data I see from a stats show -p flexscale-access into real numbers, like reduction of latency?

5 6

formatting output from 8.3 CLI statistics commands?
by Rue, Randy 07 Jul '15

07 Jul '15

I'm trying to call the "statistics volume show" command and parse it for the latency numbers for each volume. The CDOT command is trying to do some clever line wrapping so that lines with long fields break to the next line. I could just take the last field of each line but once in a while even that won't work, in some cases the bytes read is the last field of the line and the latency is the first (and last) field of the line following. Gah. Doesn't appear to be any sort of "-csv" flag for the command, or any real formatting options at all. Even tried noodling with the terminal settings used by the SSH call , thinking the line wrap might have to do with that. Can't reliably make the output behave. Anybody else faced this? Any guidance? Randy in Seattle

3 3

RE: FAS2xxx Upgrade - Flash Cache Performance Estimation?
by Stephen Stocke 07 Jul '15

07 Jul '15

Hi Toasters Sorry it's taken so long to get reply back... I've been distracted with other "fires" for the last week. A big thank you to Michael, Andrei and Tony for your responses. Toasters is a great resource because people like you are willing to share your experience and expertise. Cheers. To close this out and for the benefit of others, here's where I've landed based on your feedback: __ FlashCache vs FlashPool __ I've had some pricing back from Netapp and, as Michael said, there's no cost savings in choosing FlashCache over FlashPool for the configuration we are interested in. On that basis we'll stick with FlashPool as it's tried and tested on our busiest aggregate. It sounds like FlashPool is the future anyway so this is an easy decision. __ Flash[XXXX] Sizing __ Our oldest system, the FAS2020, doesn't support anything beyond DOT 7.3.7 so AWA is out (need 8.2 or higher) and PCS isn't supported on the platform. We just don't have any good tools available to measure the expected benefits of flash acceleration for this small system's workload. Given that the current SATA-only aggregates are performing adequately, we know the new SAS+SSD aggregates will blow our hair back. We'll buy as many of the largest SSDs our budget will support knowing that we'll get a big benefit, if not the optimum one. We also know we'll have the tool we need for sizing at the next hardware refresh (AWA). Thanks again, Steve ________________________________________ From: Michael Bergman Sent: 30 June 2015 18:21 To: Stephen Stocke Subject: Re: FAS2xxx Upgrade - Flash Cache Performance Estimation? Stephen Stocke wrote: > When planning an upgrade for a FAS2000 series filer, how do you size the > required FlashCache or FlashPool storage for the workload without having PCS > or AWA available? I read up on AWA just now. I cannot find anything in the FlashPool Tech FAQ (May 2014) that points to 'wafl awa' not being available in FAS2000 boxes. From 8.2.1 and up, in both 7-mode and c-mode. It doesn't consume much resources from the system at all: AWA uses 2% or less CPU, regardless of the number of aggregates that are simultaneously under analysis, and it uses less than 0.2% of system memory per aggregate. Even in a smaller FAS2000 model, this should be ok IMO. As I mentioned before, you really want 8.3.1 with HyA (FlashPool) on your new 8020 :-) Among other things, this limitation is changed, perhaps even removed: "Only random writes, or more precisely overwrites, that are 16KB or smaller are candidates to be cached by Flash Pool." As far as I'm aware there's also some new tuning (policy) which is the equiv of this for FlashPool: flexscale.lopri_blocks on (same value in local+partner recommended) It caches large(r) sequential reads. As it is with HyA in 8.3 it cannot be made to do that, so all seq READ comes from spinning disk. This can clearly be seen e.g. when baselining a backup with lots of SnapVault jobs running in parallel. /M -------------------------------------------------------------------------------- This e-mail message has been scanned for Viruses and Content and cleared by MailMarshal -------------------------------------------------------------------------------- disclaimer This email communication does not create or vary any contractual relationship between LogicalisCI and you. Internet communications are not secure and accordingly LogicalisCI does not accept any legal liability for the contents of this message. The contents of this email are confidential to the intended recipient at the email address to which it has been addressed. It may not be disclosed to or used by anyone other than this addressee, nor may it be copied in any way. If received in error, please contact LogicalisCI on the above number (listed as Main) quoting the name of the sender and the addressee and then delete it from your system. Please note that neither LogicalisCI nor the sender accepts any responsibility for viruses and it is your responsibility to scan the email and attachments (if any). Please be aware that LogicalisCI may monitor email traffic data and also email content for security purposes. The following companies are collectively referred to in the above statements as LogicalisCI: Logicalis Channel Islands Limited, Registered in Jersey No: 67161, Registered Office: Rue A La Dame, Five Oaks, St. Saviour, Jersey, JE2 7NH Logicalis Jersey Limited, Registered in Jersey No: 10124, Registered Office: Rue A La Dame, Five Oaks, St. Saviour, Jersey, JE2 7NH Logicalis Guernsey Limited, Registered in Guernsey No: 10896, Registered Office: Pitronnerie Road, St. Peter Port, Guernsey, GY1 2RF

2 1

alerting for faulty disks
by Caroline Yates 06 Jul '15

06 Jul '15

Hi, Last week I happened to notice a disk which was "disabled due to excessive phy changes". I contacted Netapp support and they promptly sent a replacement. Of course I asked why this didn't show up in the failed disk count and they replied that it was "faulty not failed" (although the message in the logs say "Call home for DISK_FAIL - Disabled by IOM"). Anyway, my concern of course is to catch these faulty disks. My preference for monitoring is Nagios and snmp but having snmpwalk-ed through the current mib (ontap 8.1.1) I don't see an oid that would capture this faulty disk status - unless this is considered "prefailed"? But I am more concerned with catching this problem than I am wedded to Nagios, so any advice would be appreciated. Thanks! Caroline ______________________________________________________________________ Caroline Yates, Unix Systems Contractor, CYates(a)slb.com<mailto:CYates@slb.com> Schlumberger, High Cross, Madingley Road, Cambridge, CB3 0EL Tel: +44 (0)1223 325276 Mob: +44 (0)7753 353889 ______________________________________________________________________ Registered Name: Schlumberger Cambridge Research Limited Registered Office: Schlumberger House, Buckingham Gate, Gatwick Airport, West Sussex. RH6 0NZ, United Kingdom Registered in England and Wales No. 00606177 Note: The Gatwick airport address is NOT the street address of the Cambridge site., this is postcode CB3 0EL

1 0

SVM DR in 8.3.1
by Milazzo Giacomo 06 Jul '15

06 Jul '15

Hi all, I know is still RC1 but prior to upgrade a running 8.3 I would try to have more info on a new 8.3.1 feature: the SVM DR. I've a customer that on 7Mode ran the DR of CIFS vfiler (joined to a domain) using snapmirror of the vfiler vols, including the root one, and at the DR test/run the source vfiler was stopped and a vfiler create -d command, with the mirrored root specified, let the destination vfiler start with all domain infos, shares on the vole and ip addresses on the DR site with minimal effort. I've read the docs that come with 8.3.1 but is not clear. I don't think that there's something in SVM DR that permit to recreate the SVM on the destinanation: this must exist in advance, in stopped mode, right? What's about the maintenance of the IPs? Should I have two identical subnets in both clusters? What if I cannot have two different subnets? Do I have a way to let the replicated SVM being in standby with the data lif with preregistered IPs? Regards

4 7

cDOT and Oracle
by Milazzo Giacomo 06 Jul '15

06 Jul '15

Hi everybody, one million dollar questions ;-) aka transition to cDot could be really a nightmare! The questions come not only from best practice and TR I've read (or I could read...) but your ideas about best ways to bring in cDot what now is deeply optimized in 7Mode. AS is - 7 Mode A lot of NFS volumes where to separate data/log/temp of each Oracle instance. Each volume is exported for access only to the hosts that manage binaries and db. In the production environment all instances are RAC while in development test most of them are RAC but some are also composed just by one server. 90% of hosts are VMware VM, some are phyical for RAC instances composed by a phys-virtual pair of hosts. Volumes are hosted on aggregates composed by SAS or FC discs depending on type and importance of workload. In test/dev there are mostly SATA vols but non only. Of course, all the instances are managed and replicated with the concurrent usage of Snapcreator and its integration with DFM Protection Manager that it's then integrated with Snapmirror To be - cDOT (8.3.1) <- we hope to be able to use SVM DR features somewhere... Each instance in a its own SVM? This could be a nightmare to manage ipspaces, dedicated lifs, flexvol that once assigned are locked to that SVM and so on.... Just one SVM for all NFS RAC? Maybe creating differente FlexVols with qtree for data/log/temp? Or on different aggregates? One or more SVM, few, just to balance the workloads between controllers? Is there an advantage offered by the juction paths that should be created? Every idea is accepted also is recently has been demonstrated that brainstorming is useless :) Regards, Dott. Giacomo Milazzo Senior Consultant & Technical Account Manager mobile: +39 340.6001045 @-mail: g.milazzo(a)sinergy.it<mailto:g.milazzo@sinergy.it> Web: http://www.sinergy.it<http://www.sinergy.it/> [cid:4DB1B876-8A76-4829-9B40-9B136E344F73@fastwebnet.it] SINERGY SpA Viale dei Santi Pietro e Paolo 50 00144 - Roma RM Tel. +39 06 44243674 Fax +39 06 44245272

2 2

NFS & TR4067
by Mark Flint 06 Jul '15

06 Jul '15

Think this may be a version specific issue, but I’m trying to get client stats out of an 8.2P6 CDOT system, preferably sorted so that I can see the top few clients. I’ve found the above TR by Justin (thanks :) ) but one of the useful commands seems not to work. I’m using “statistics settings modify -client-stats enabled” and it returns a not found, only reference is that the statistics setting modify command can only change the counter display rate :( Must be a way to do this, or do I have to try and write something? Got a bit of help from Douglas’ comments here yesterday, but still struggling…. ~Mark mark.flint(a)sanger.ac.uk

3 5

Re: Another SSD question
by Basil 05 Jul '15

05 Jul '15

I couldn't find anything about CIFS and Flash Pool not working together- could you send me a link? Flash Pool is at the aggregate level and shouldn't care what kind of data is in the volumes. On Sat, Jul 4, 2015 at 7:34 PM, Neil Stichbury <Neil.Stichbury(a)spark.co.nz> wrote: > As FlashPool & CIFS are (currently) mutually exclusive I can only see > FlashCache in CIFS environments in the future. The Netapp FlashPool sizing > suggestions are a good place to start and you can start small, keep an eye > on the stats and avoid diminishing returns. cDOT 8.3 supports partitions > and adding partitions is fine though removing them is not somewhere you > want to go! > > > > Speaking of future, the point at which it becomes more expensive to > assemble and provide a spinning disk assembly than a similar high-capacity > SSD will be an overnight game-changer consigning these arguments to the > archives. Any predictions on year/month? Sweepstake anyone? I’m not > sure that the industry really wants to see this happen, so make take a > while ;-) > > Neil > > > > *From:* toasters-bounces(a)teaparty.net [mailto: > toasters-bounces(a)teaparty.net] *On Behalf Of *Basil > *Sent:* Saturday, 4 July 2015 7:31 a.m. > *To:* Tony Bar > *Cc:* toasters(a)teaparty.net > *Subject:* Re: Another SSD question > > > > Thanks for the note about the card slots- I'm checking to see whether we > even have a spare. Also, that wafl awa command is precisely what I was > looking for, thanks! > > > > On Fri, Jul 3, 2015 at 2:57 PM, Tony Bar <tbar(a)berkcom.com> wrote: > > Basil - > > > > I would be very careful with choosing in this case. FlexPool may be > marginally more expensive but seeing as how the 8020 only has two PCIe > slots adding a FlashCache module will severely limit your expansion > options. If that other slot is occupied by another 10GbE card you'll be > limited to one SAS stack (10 shelves). Furthermore, per best practices, > mixing SAS/SATA drive types on the same stack is not encouraged (although > it can be done), so you're essentially limiting yourself to one drive > type. On the other hand, if the other slot is occupied by a SAS card then > you're limiting your connectivity options. If you're running CDOT you're > already going to be giving up two 10GbE onboard ports for the cluster > interconnect so effectively you have two 10GbE/FC ports per controller. If > in the future you decided you wanted to add a tape drive (FC) or more 10GbE > ports you'd be out of luck. > > > > Regarding performance, PCS doesn't really differentiate, and to be honest > for most real world situations I've seen it's a wash. The ability to have > write caching via FlashPool is very helpful for random overwrite workloads. > 'wafl awa' can help you determine the impact adding cache of either kind > will have on writes/reads and how much optimally you would need. It is > important to run it at a peak time to see what the recommendations are, as > running it during a relatively low activity period won't give much useful > information. > > Anthony Bar > 650.207.5368 > > tbar(a)berkcom.com > > > On Jul 3, 2015, at 10:03 AM, Basil <basilberntsen(a)gmail.com> wrote: > > So I'm looking at upgrading a few 8020s and trying to choose between > flashpool and flashcache. Flashpool is more expensive, even with just > 12x200 SSD drives, and we'd need to find space in a very tight rack for it. > The workload is 80% exchange and 20% VMWare, and we would like to use > deduplication without snapshots. > > > > Is it possible to use PCS to compare flashcache and flashpool benefits? > How can I translate the data I see from a stats show -p flexscale-access > into real numbers, like reduction of latency? > > _______________________________________________ > Toasters mailing list > Toasters(a)teaparty.net > http://www.teaparty.net/mailman/listinfo/toasters > > >

2 1

v3170 Setup
by Chris hansen 05 Jul '15

05 Jul '15

I have a old v3170 in my lab with one MK2 SATA shelve. I am trying to find documentation on setup for wiring and HA setup. I have one side working but can’t seem to get both sides working….assume I have some wiring wrong. Searching online I can’t find much out on this model. Thanks -Chris

3 2

Quickest way to format
by Camillo Gornati 03 Jul '15

03 Jul '15

Hello All! Via console, what is the quickest, easiest and safest way to destroy an aggregate and CIFS. Basically the best way to bring a FAS2040 to its factory condition. Thanks

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