toasters

toasters@lists.teaparty.net

3 participants
13514 discussions

Is there a limit to the number of VM's per NFS export?
by Page, Jeremy 28 May '08

28 May '08

For the folks using NFS to hold their VMDK files, have you run into a limit to how many VM's a single share can support? Either a hard limit or just a point where performance degrades? I'm trying to figure out how many volumes I'll need to create to hold ~300 VMs on my 3070 cluster. 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.

4 5

"df -h" shows less snapshots than "snap delta"
by Borzenkov, Andrey 27 May '08

27 May '08

I am rather puzzled. cn1:~ # rsh filer df -h vol1 Filesystem total used avail capacity Mounted on /vol/vol1/ 2448GB 776GB 1671GB 32% /vol/vol1/ snap reserve 612GB 158GB 453GB 26% /vol/vol1/.. So according to this we have 156GB worth of snapshots. So far so good. cn1:~ # rsh filer snap delta vol1 Volume vol1 working... ... Summary... From Snapshot To KB changed Time Rate (KB/hour) --------------- -------------------- ----------- ------------ -------------- Oldest_snap Active File System 265800060 14d 09:08 770120.583 Oops. I would expect df -h to show *more* than snap delta (due to the fact that we have more intermediate snapshots); but how comes that - apparently having 260GB worth of snapshot data - I only see 158GB in space accounting? I am likely missing something obvious. Hmm ... is it possible that "snap delta" accounts for both changed *and* new data while "df -h" accounts for *changed* data only? Looks plausible... С уважением / With best regards / Mit freundlichen Grüβen --- Andrey Borzenkov Senior system engineer

3 3

RE: NetApp & Leopard
by Villabroza, Gerald 24 May '08

24 May '08

Patrick, Tough to mandate dave or admitmac in a diverse higher education environment. 100's of macs show up after the Christmas holidays and they all expect to use university resources immediately. Carl, Our understanding from Apple is that the next Leopard update, 10.52, will address the CIFS access issue. It's in a testing phase now but not available to folks external to Apple. -=-=- gerald villabroza <geraldv at stanford.edu> technical lead, its storage, stanford university > -----Original Message----- > From: Patrick van Helden [mailto:pvh@databasement.eu] > Sent: Wednesday, January 30, 2008 8:24 AM > To: Carl Howell; Villabroza, Gerald > Cc: toasters(a)mathworks.com > Subject: RE: NetApp & Leopard > > Hi Guys, > > Why don't you guys use a 3rd party client like "Dave" or "Admitmac" > from Thursby? > > Admitmac even has Windows DFS support > > Regards, > > Patrick van Helden > Databasement BV > pvh(a)databasement.eu > > > > -----Oorspronkelijk bericht----- > Van: owner-toasters(a)mathworks.com namens Carl Howell > Verzonden: wo 1/30/2008 15:56 > Aan: geraldv(a)stanford.edu > CC: toasters(a)mathworks.com > Onderwerp: RE: NetApp & Leopard > > Gerald, > > Thanks for the feedback, and yes, feel free to reference us. > > --Carl > > -----Original Message----- > From: Villabroza, Gerald [mailto:geraldv@stanford.edu] > Sent: Wednesday, January 30, 2008 8:49 AM > To: Carl Howell > Cc: toasters(a)mathworks.com > Subject: Re: NetApp & Leopard > > Carl, > > We're experiencing the same issue when accessing DOT 7.2.2 CIFS in Win > 2k3 AD with OS X 10.5.1. > > We've opened a case with Apple and here's what they came back with: > > ##### > When a Leopard client opens a session, it sends three mechanisms in > this > > order, KRB5, some OID I don't what it is, and MS KRB5. The filer > returns an unsupported error. > > Apple thinks DOT is just bailing on the first unsupported mechanism and > not checking the whole list. Tiger only sent the MS KRB5 mechanism so > that is why it works. > > Apple is working on building a test of their kerberos library that puts > MS KRB5 as the first mechanism to validate the hypothesis. > ##### > > Leopard can authenticate via K5 against MS WIN 2k3 systems fine in our > environment, just not against DOT. > > Luckily Apple and NetApp are both TSAnet members and can collaborate on > the support case. > > Do you mind if reference your experience at UWF with NetApp and Apple? > And if you don't, do you have a case # with NetApp? > > Its interesting to hear of other hi-ed's with this issue. Any others > out there? Like other issues in our space it helps to band together. > > -=-=- > gerald villabroza <geraldv at stanford.edu> > technical lead, its storage, stanford university > > > Carl Howell wrote: > > I've stumbled across a problem we're having accessing filer hosted > CIFS > > shares from Mac OS X Leopard 10.5.1. The Leopard boxes I've tried > this > > on are all bound to our Win2k3 Active Directory. If you log into > Leopard > > with your domain credentials and try to access a share on a > filer(this > > happens on all of our filers and all are at 7.x and above), you will > be > > prompted for your password. If you try to access the same CIFS share > > hosted on a Win2k3 box, you will get right in. > > > > > > > > Has anyone else seen this? > > > > > > > > Thanks, > > > > > > > > --Carl > > > > > > > > > > > > > > >

4 5

RE: 10 Gbps curiosity
by Glenn Walker 22 May '08

22 May '08

Crap :-( NetApp - when are you going to enable ToE and EtherChannel together? This bombs my plans pretty quickly. Maybe I can change it now before it's too late. ________________________________ From: Myles Uyema [mailto:mlists@uyema.net] Sent: Wednesday, May 21, 2008 10:21 PM To: Glenn Walker Cc: Jean-Noel Filippi; Milazzo Giacomo; <toasters(a)mathworks.com> Subject: Re: 10 Gbps curiosity With TOE disabled we maxxed out a 6070 head at 375MBytes per second on 7.2.4. NFS only. Network domain pegged at 100%. With TOE enabled on 7.3RC1 we saw 815MBytes per second. 60% network domain. PCI-X single port TOE, 100 clients reading sequentially. </phone> On May 21, 2008, at 2:58 PM, "Glenn Walker" <ggwalker(a)mindspring.com> wrote: We're deploying FAS6070s with 2 x Active\ 2 x Passive 10GbE Single Mode, multi-level VIFs (Etherchannel) per head for VMWare over NFS. Looks good so far, so I'll let the group know how this turns out... FWIW, I have the same concerns over CPU, but I'm hedging my bets that the FAS6070s are such beasts that they won't really care. ________________________________ From: owner-toasters(a)mathworks.com [mailto:owner-toasters@mathworks.com] On Behalf Of Jean-Noel Filippi Sent: Wednesday, May 21, 2008 2:18 PM To: Milazzo Giacomo; <mailto:toasters@mathworks.com> toasters(a)mathworks.com Subject: RE: 10 Gbps curiosity We use 2 x FAS6080 with redundant 10GE in a VIF single mode. They are connected to Force10 C300 switches. We generated up to 4Gbit of CIFS traffic and everything works well. ________________________________ From: owner-toasters(a)mathworks.com [mailto:owner-toasters@mathworks.com] On Behalf Of Milazzo Giacomo Sent: Wednesday, May 21, 2008 8:11 AM To: <mailto:toasters@mathworks.com> toasters(a)mathworks.com Subject: 10 Gbps curiosity Hi all Reading on superior results reached in certaing environment (NetApp NFS with Oracle ad example) I'm looking for some direct experience and informations about performances, benchmarks, infrastructure used - I mean also exactly model and brand of switches :-) - and so on. So that, more on the relatively new world of 10 Gbps networks. Does anyone direct experience? Are you using it in your site? Thanks in advance

2 3

Re: "df -h" shows less snapshots than "snap delta"
by Webster, Stetson 22 May '08

22 May '08

Try "aggr show_space" ----- Original Message ----- From: Stephen C. Losen <scl(a)sasha.acc.virginia.edu> To: Borzenkov, Andrey <Andrey.Borzenkov(a)fujitsu-siemens.com> Cc: toasters(a)mathworks.com <toasters(a)mathworks.com> Sent: Thu May 22 07:29:53 2008 Subject: Re: "df -h" shows less snapshots than "snap delta" > I am rather puzzled. > > cn1:~ # rsh filer df -h vol1 > Filesystem total used avail capacity Mounted on > /vol/vol1/ 2448GB 776GB 1671GB 32% /vol/vol1/ > snap reserve 612GB 158GB 453GB 26% /vol/vol1/.. > > So according to this we have 156GB worth of snapshots. So far so good. > > cn1:~ # rsh filer snap delta vol1 > > Volume vol1 > working... > > ... > > Summary... > > From Snapshot To KB changed Time Rate (KB/hour) > --------------- -------------------- ----------- ------------ -------------- > Oldest_snap Active File System 265800060 14d 09:08 770120.583 > > Oops. I would expect df -h to show *more* than snap delta (due to the fact that we have more intermediate snapshots); but how comes that - apparently having 260GB worth of snapshot data - I only see 158GB in space accounting? > > I am likely missing something obvious. Hmm ... is it possible that "snap delta" accounts for both changed *and* new data while "df -h" accounts for *changed* data only? Looks plausible... Snapshot accounting can be very tricky I think. Consider this scenario. 1) take snapshot #1 of a volume 2) create a new and very large file 3) take snapshot #2 of volume 4) delete the very large file 5) take snapshot #3 At this point the "delta" between snapshots 1 and 3 and the live volume will be very small. But the "delta" between snapshot 2 and any other snapshot (or the live volume) will be large. This is because snapshot 2 (and only snapshot 2) still contains the large file that was deleted. So you can see that the delta between the live volume and each of its snapshots does not necessarily increase with the age of the snapshot. Steve Losen scl(a)virginia.edu phone: 434-924-0640 University of Virginia ITC Unix Support

1 0

Snapmanager for Virtual Infrastructure
by Milazzo Giacomo 22 May '08

22 May '08

Hi all Among (poor) white papers, announcements and so on, Snapmanager for VMWare (or VI3) has been announced on february 2008... Is there somebody explain why I can't find detailed white papers, tecg resosurce docs on now and so on? And, more, is there somebody who tested it and that can tell me (us) if this produtct can really do more that the script we usually use to backup VMs on NetApp storage? Thanks a lot

12 20

10 Gbps curiosity
by Milazzo Giacomo 21 May '08

21 May '08

Hi all Reading on superior results reached in certaing environment (NetApp NFS with Oracle ad example) I'm looking for some direct experience and informations about performances, benchmarks, infrastructure used - I mean also exactly model and brand of switches J - and so on. So that, more on the relatively new world of 10 Gbps networks. Does anyone direct experience? Are you using it in your site? Thanks in advance

5 5

Benfits of using LACP over Etherchannel?
by Page, Jeremy 20 May '08

20 May '08

I have a two VIFs (one for each filer head) spread across a pair of stacked 3750-Es. Right now I'm using Etherchannel on the switches but I've been reading a lot saying that LACP may be better. I'm trying to figure out what exactly the benefit would be, can anyone clarify? 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.

2 1

Re: R: MSCS and snapmirror question
by Klise, Steve 19 May '08

19 May '08

Correct. Same filer. Thanks ----- Original Message ----- From: Milazzo Giacomo <G.Milazzo(a)sinergy.it> To: Patrick van Helden <pvh(a)databasement.eu>; Klise, Steve; toasters(a)mathworks.com <toasters(a)mathworks.com> Sent: Fri May 16 10:29:55 2008 Subject: R: MSCS and snapmirror question Of course but that guy said he's migrating on the same filer to another aggregate. ________________________________ Da: Patrick van Helden [mailto:pvh@databasement.eu] Inviato: ven 16/05/2008 19.12 A: Milazzo Giacomo; Klise, Steve; toasters(a)mathworks.com Oggetto: RE: MSCS and snapmirror question Guys, Please be aware that MSCS will keep track of the lun serials. So you would probably need to set the lun serial numbers back to the original ones when your snapmirroring to another filer. Regards, Patrick van Helden Databasement BV pvh(a)databasement.eu -----Oorspronkelijk bericht----- Van: owner-toasters(a)mathworks.com namens Milazzo Giacomo Verzonden: vr 5/16/2008 18:01 Aan: Klise, Steve; toasters(a)mathworks.com Onderwerp: R: MSCS and snapmirror question You will not have issues. And you don't need to update your mirror. If you simply power off you clusters no i/o activity will be performer on LUNs. Then you can simply copy (vol copy) the volume(s) containing your LUNs, rename the new volumes as the old (in case of iSCSI LUNs and if you don't want to recreate the connections) and restart you clusters. In case your nodes (but it will not happen) will loose the LUNs with Snapdrive you can safely reconnect them starting with the quorum one, on first node then restarting the cluster service, then on the second node and so on with all other cluster shared LUNs Regards -----Messaggio originale----- Da: owner-toasters(a)mathworks.com [mailto:owner-toasters@mathworks.com] Per conto di Klise, Steve Inviato: venerdì 16 maggio 2008 15.50 A: toasters(a)mathworks.com Oggetto: MSCS and snapmirror question I am migrating from an aggregate to another aggregate in the same filer (6030). The LUNS are attached to print clusters, 2 node and 3 node clusters (WIndows 2003 sp2 x86 etc). I have snapmirrored the volumes. If I stop the cluster service on the windows boxes, do a snapmirror update, the disconnected the quorum and other resource disks, do a snapmirror update, release the snapmirror, remove and replace the disks with snapdrive with the same disk letters. fire up the cluster service.. Will that work? has anyone done that before? Any gotchas? Any help is greatly appreciate!

3 4

Re: Oddball SnapMirror issue
by Webster, Stetson 19 May '08

19 May '08

Volume Snapmirror, or Qtree Smapmirror? ----- Original Message ----- From: Raj Patel <phigmov(a)gmail.com> To: Kevin Graham <kgraham(a)industrial-marshmallow.com> Cc: toasters(a)mathworks.com <toasters(a)mathworks.com> Sent: Sun May 18 15:54:47 2008 Subject: Re: Oddball SnapMirror issue > Try backing off SnapMirror's TCP window (options snapmirror.window_size) > to lower the effective throughput, or nail down 'kbs' options in your > snapmirror.conf. Yup - tried the "options snapmirror.window_size 32768" and throttling the mirror speed down. I still don't understand why, if the connection was ropey, it didn't just keep trying to establish the baseline rather than just giving up and failing to initialise. I'll have to see what our networking guys come up with on the ipsec side. Cheers, Raj.

2 1

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