toasters

toasters@lists.teaparty.net

13520 discussions

RE: C720 things
by niall.doherty＠netapp.com 25 Jul '00

25 Jul '00

> Thanks for that Niall, interesting.. I'm very keen to do > it, if you want to try a beta version of the software in > a production environment then let me know... We can definitely keep you in mind ! :-) Thanks ! > I know theres currently options on the netapp to allow > spoofing, Yes, the support needed on the cache server side is not too hard. Just accepting packets for us that were actually destined to a client instead ;-) (returning traffic) > but your document suggests this isnt actually workable > at the moment as the l4 switches dont redirect the return > traffic.. is that right? Woah. I said Foundry can't do it (yet), but Alteon can. We have customers doing this already. Note that you need to make sure that your traffic is "symmetrical" - i.e., the returning traffic will definitely go through the same device - otherwise you'll have problems. You have to be *very* careful with this feature... > so right now theres no way to use my cache and maintain > the client ip throughout the connection? With Alteon you can - right now. With WCCP 2 you will be able to in the (near?) future. And Foundry will hopefully add the support soon too... [ As you probably know, NetCache adds an X-Forwarded-For header (if none already exists) to all requests it sees. This contains the client IP address. So web servers can get access to it that way if they want to... ] Cheers, Niall

1 0

Monitoring other stuff with MRTG
by Henrique Pantarotto 25 Jul '00

25 Jul '00

Hello Toaster friends, I am already monitoring my filers for these infos: - CPU utilization - NFS ops/seg - /vol/vol0 volume - ethernet traffic I was wondering if its possible to also monitor: - Cache age (as reported by sysstat command) - Disk kB/s read and write (also reported by sysstat) Any know know the SNMP OID to this? I tried searching at the MIB, but I couldn't find it. Thanks, _______________________________________________ Henrique Pantarotto SysOp Site São Paulo Terra Networks Brasil S/A A Internet mais sua do que nunca Tel: (11) 5505-5728 r.316/238 ICQ: 6934285 IT: henpa henrique(a)corp.terra.com.br

1 0

RE: C720 things
by niall.doherty＠netapp.com 25 Jul '00

25 Jul '00

> On -another- thread, I'm running WCCP and of course all web > requests show the IP on the cache not the client itself.. is > there a way to spoof the client's ip? I dont have a L4 switch > btw, i run WCCP from a CAT6000 with MSFC. I have attached some info I wrote up a few days describing IP spoofing. WCCP 2 is mentioned in here. To summarise, it should be possible with WCCP 2 but we haven't tested it or added the necessary support just yet. It'll be available "soon". Coding it should be straightfoward but there are other release issues to work on first... Cheers, Niall -- Niall Doherty | Network Appliance Inc. | NetCache Deployment Architect | 495 East Java Drive | Intercontinental Operations | Sunnyvale, CA 94089, USA | mailto:ndoherty@netapp.com | http://www.netapp.com | Direct Line: +1 408 822 8554 | HQ Phone: +1 408 822 6000 | FAX(->email): +1 801 912 5212 | HQ FAX: +1 408 822 6001 |

2 1

Re: Binary numbering (off topic)
by mds＠gbnet.net 25 Jul '00

25 Jul '00

On Tue 25 Jul, 2000, Jay Orr <orrjl(a)stl.nexen.com> wrote: > > I had brought up the question a few months back of if 1GB is 1000KB or > 1024KB. Yes. Both, as the de jure and de facto standards are different. > Now as everyone knows, historically 1024 Bytes is 1KB. Seems marketing has > won the battle in terms of calling a "kilo" 1000 and not 2^10, at least > with the government. I have been involved in a thread on > comp.sys.sun.hardware about this topic and thought I'd share a posting > about this. I'm hoping I'm not the only one unaware of this "new" > standardization... Yeah, I've also heard the case presented by the faction that proposes Mibibyte, Mibibit, Kibibyte, Kibibit, etc. (sp for all those) to denote the 2^10^3n quantities. Seems quite reasonable to me.. except in the face of enormous socio-linguistic inertia I'm not sure I'd fight that hard. It's very nearly always been denary powers in raw disk sizes. I still think the binary powers are firmly ensconced in RAM, file sizes and bandwidth (MHz -> Mb and MB seems to pose less problem for people headers and preambles notwithstanding) in the mind of both the novice and the old-hand. More same-old same-old I think. I'll continue using whatever units my listeners/readers will understand most readily - anything else is poor communication. The issue that causes _me_ the most irritation is when people insist on using lowercase 'b' in Kb, Mb, etc. when they really mean 'byte' which, for me, and I suspect for many of us, would suggest 'bit'. Or when they're inconsistent. Sometimes it's not always possible to deduce from the context what they really mean, and, worse yet, they often don't see the issue as a problem! > Camiel Vanderhoeven wrote: > > > > Yes, the Ki, Mi, Gi etc. prefixes are a nice initiative to avoid > confusion. > > However, I have never seen anyone who seriously uses it. Old habits > don't > > vanish that easily. I think the simbol "KB" and the word "kilobyte" have > been > > used for 1024 bytes since the early days of computing. You can't expect > > everyone to change this into "KiB" and "kibibyte" because the SI > committe > > decides so in 1998. Not seriously. > > > > Camiel. > > > > hac <hchristeller(a)home.com> wrote: > > > > > Jay Orr wrote: > > > > > > > > I got into a discussion about a similar matter on another discussion > > > > group. Some vendors think 1 GB is 1000 MB, while others will say 1 > GB is > > > > 1024 MB. Varies by OS too. No one could is quite sure who uses > which. > > > > So, it could also be a 650MB CDROM would hold 0.634 GB.... > > > > > > > 1 GB is 1000 MB, 1 GiB is 1024 MiB. Any other use is sloppiness. > > > > > > http://physics.nist.gov/cuu/Units/binary.html > > > > > > -- > > ----------- > Jay Orr > Systems Administrator > Fujitsu Nexion Inc. > St. Louis, MO > >-- End of excerpt from Jay Orr -- -Mark ... an Englishman in London ...

1 0

Quotas for NT users
by Ferdie Mazon 25 Jul '00

25 Jul '00

Question of the day: What is the best way to use quotas in an NT environment? My user list changes frequently, and my groups are a mess. Thanks, Ferdie

1 0

Binary numbering (off topic)
by Jay Orr 25 Jul '00

25 Jul '00

I had brought up the question a few months back of if 1GB is 1000KB or 1024KB. Now as everyone knows, historically 1024 Bytes is 1KB. Seems marketing has won the battle in terms of calling a "kilo" 1000 and not 2^10, at least with the government. I have been involved in a thread on comp.sys.sun.hardware about this topic and thought I'd share a posting about this. I'm hoping I'm not the only one unaware of this "new" standardization... Camiel Vanderhoeven wrote: > > Yes, the Ki, Mi, Gi etc. prefixes are a nice initiative to avoid confusion. > However, I have never seen anyone who seriously uses it. Old habits don't > vanish that easily. I think the simbol "KB" and the word "kilobyte" have been > used for 1024 bytes since the early days of computing. You can't expect > everyone to change this into "KiB" and "kibibyte" because the SI committe > decides so in 1998. Not seriously. > > Camiel. > > hac <hchristeller(a)home.com> wrote: > > > Jay Orr wrote: > > > > > > I got into a discussion about a similar matter on another discussion > > > group. Some vendors think 1 GB is 1000 MB, while others will say 1 GB is > > > 1024 MB. Varies by OS too. No one could is quite sure who uses which. > > > So, it could also be a 650MB CDROM would hold 0.634 GB.... > > > > > 1 GB is 1000 MB, 1 GiB is 1024 MiB. Any other use is sloppiness. > > > > http://physics.nist.gov/cuu/Units/binary.html > > > > -- ----------- Jay Orr Systems Administrator Fujitsu Nexion Inc. St. Louis, MO

1 0

RE: Ndmp vs. Nfs backups.
by Kelly Wyatt 25 Jul '00

25 Jul '00

We use DLT7000, the compression devices on the netapp. Right now the master for these backups is a sun ultra 2. He really does not do much other than schedule jobs, perform database updates, etc. He has to create the list of files to backup and has to have a lot of temporary disk space and memory to handle this task. During the backups, most (all?) of the CPU cycles come from the filers -- all 760's. Here is some info on our sun: Sun Ultra 2 UPA/SBus (UltraSPARC-II 296MHz) with 384MB of memory. We have a 70GB raid array for /usr/openv to allow space for database growth. I thought he had more memory it may have been "borrowed" for an unexpected project. We do stagger our backups so that all backups jobs are not going for resources at one time. Each filer volume is a class for a lot of reasons. Originally we tried to limit our volumes to a size that can be restored within a 4 hour window... in reality, it is probably closer to 6 hours on some volumes, and we have a couple of "legacy" volumes from pre-multi volume days that are about 150GB. We have tested restores and seem to get around 15-17gb/hour restoring data, depending on the data. Kelly -- Kelly Wyatt, Kelly.Wyatt(a)SAS.com Senior Systems Programmer Strategic Recovery SAS Institute Inc. / SAS Campus Drive / Cary, NC 27513 http://www.sas.com SAS... The Power to Know -----Original Message----- From: Jay Orr [mailto:orrjl@stl.nexen.com] Sent: Tuesday, July 25, 2000 9:16 AM To: Kelly Wyatt Cc: 'Brian Hostetter'; toasters(a)mathworks.com Subject: RE: Ndmp vs. Nfs backups. Thanks for the data, it is quite useful! What kind of storage medium are you using (i.e. DLT?, AIT?) and what is your backup hardware (i.e. what are the specs of the system you use to run the backup software on)? This is the kind of info I can't even get the sales 'drioids to get for me... ----------- Jay Orr Systems Administrator Fujitsu Nexion Inc. St. Louis, MO

1 0

RE: Ndmp vs. Nfs backups.
by Kelly Wyatt 25 Jul '00

25 Jul '00

I have not tested NFS mounted backups versus NDMP direct attach (we use these in a multiprotocol manner, and with NFS mounts you lose the CIFS acls), but I have done backups to remote NDMP storage units using NetBackup. I hope this data is useful. Following is a note I sent to our management earlier this year. This is using NetBackup. The data was collected from NetBackup and thrown through SAS to perform the calculations. Kelly ----- I have been wondering what the affect of local (direct attach drive) versus remote (over the network) backups is on the NetApps. The following reports show that difference... For both reports (Full backups first report, Incr backups second report), the LOCAL is the mean GB/hour for local backups and the REMOTE is the mean GB/hour for remote backups. Where there is no REMOTE it means that there have been no remote backups for that volume. The PERDIFF is the difference between the LOCAL and the REMOTE (REMOTE/LOCAL). The anomalies (where remote is > ) probably means that the first big backup was an incremental backup, which, in reality, is equivalent to a full. Also, there are some volumes that always backup to remote filers... and this is indeed correct. This is on 100ethernet. We are looking at setting up a private GB network for remote backups. Full Backups 1 OBS CLASS LOCAL REMOTE PERDIFF 1 17.7295 8.1000 0.45687 2 b.vol0 24.9750 13.1800 0.52773 3 b.vol1 . 13.3340 . 4 h.vol0 . 12.7200 . 5 h.vol1 21.5450 12.6000 0.58482 6 h.vol2 0.5800 0.4725 0.81466 7 s.vol0 0.7200 1.8567 2.57870 8 s.vol1 15.8200 15.3133 0.96797 9 s.vol2 0.0400 0.0550 1.37500 10 s.vol3 . 0.0467 . Incremental Backups 2 OBS CLASS LOCAL REMOTE PERDIFF 1 11.6673 7.2805 0.62401 2 b.vol0 20.6550 13.6800 0.66231 3 b.vol1 11.4783 4.2700 0.37201 4 e.vol0 15.4419 12.2767 0.79502 5 e.vol1 15.6693 11.7867 0.75221 6 h.vol0 13.5387 11.0275 0.81452 7 h.vol1 19.7640 11.5750 0.58566 8 h.vol2 0.1453 0.0300 0.20642 9 k.vol0 9.2976 7.4225 0.79832 10 k.vol1 9.6065 6.9500 0.72347 11 k.vol3 10.7789 10.5400 0.97784 12 o.vol0 15.2288 9.4967 0.62360 13 o.vol1 6.7781 4.2133 0.62161 14 o.vol2 10.7219 8.6333 0.80521 15 p.vol0 1.1365 0.0700 0.06159 16 p.vol1 13.5583 9.6750 0.71358 17 p.vol2 14.8133 12.4600 0.84113 18 p.vol3 14.1555 9.8500 0.69584 19 q.vol1 0.2250 0.3300 1.46667 20 r.vol4 11.1685 11.3500 1.01625 21 s.vol1 13.4850 8.9160 0.66118 22 s.vol2 3.3287 0.0333 0.01001 23 s.vol3 0.0336 0.0200 0.59459 -- Kelly Wyatt, Kelly.Wyatt(a)SAS.com Senior Systems Programmer Strategic Recovery SAS Institute Inc. / SAS Campus Drive / Cary, NC 27513 http://www.sas.com SAS... The Power to Know -----Original Message----- From: Brian Hostetter [mailto:BrianH@dice.com] Sent: Monday, July 24, 2000 12:47 PM To: toasters(a)mathworks.com Subject: Ndmp vs. Nfs backups. Has anyone tested the speeds of ndmp backups vs nfs? I am looking at three different backup packages, and all have fuctions that will not work with ndmp (archiving, tape duplication, cataloging). I am wondering if nfs is any where near the backup speed of ndmp. If it is close then the benefits may outweigh the speed loss. Right now I am getting about ~8M/s per drive with ndmp which is really nice.

2 1

RE: Netapp examples - replication for disaster recovery?
by Anders Ljungberg S (QDT) 25 Jul '00

25 Jul '00

Hello Jim! Yes, at Ericsson in Stockholm we use this setup for disaster/recovery reasons. But we also use the replicated data to take our backups from. I.e we dont have to load our "Frontend Filers" with backup traffic. We have 10 F760's in 5 clusters replicating there volumes to 5 other F760's on another location. 5-6TB of data. We do this twice an hour and see 10-20sec bursts of data going over. Anders Ljungberg Ericsson Business Consulting -----Original Message----- From: Jim Klun 7183 [mailto:jklun@stercomm.com] Sent: Tuesday, July 25, 2000 1:37 PM To: Jim Klun 7183 Cc: Toaster Mailing List Subject: Re: Netapp examples - replication for disaster recovery? Thanks for responses - I noted that there was little mention of replication. Does anyone out there use replication filer-to-filer for disaster recovery or fail-over? I would like filers at sister sites replicating to one another to allow faster disaster recovery. Does anyone do this? * James A. Klun Sterling Commerce 4600 Lakehurst Ct. Dublin, Oh 43016 * * jklun(a)stercomm.com 614-793-7183 voice 614-793-7092 fax * * ---- "very like a whale" ---- * * ---- Hamlet, Act 2, Scene 2 ---- * On Mon, 24 Jul 2000, Jim Klun 7183 wrote: > > > I want to explain the NAS concept as implemented via Netapps in an > upcoming meeting. > > I am looking for real world examples where snapshot saved someones lunch. > Also uses of volume replication for 7X24 and DR. > > Finally - advantages - if they exist - in running Microsoft Exchange mail > off a netatpp. > > Thanks - meetings tomorrow. > > > > > * James A. Klun Sterling Commerce 4600 Lakehurst Ct. Dublin, Oh 43016 * > * jklun(a)stercomm.com 614-793-7183 voice 614-793-7092 fax * > * ---- "very like a whale" ---- * > * ---- Hamlet, Act 2, Scene 2 ---- * > >

1 0

C720 things
by Stephen J. Wilcox 25 Jul '00

25 Jul '00

Hi, it seems the C720 software has a filelimit of 2.2M objects imposed upon it, apparently netapp are due to increase this limit in a few months but in the meantime my C720 is using less than half its drive space so I want to get the max usage! At the moment I've got 51% object hit rate and 26% byte rate.. not too bad but I suspect I can increase it. As I understand it, when the file limit is hit if it retrieves a new object it replaces one of the older objects on the disk. Now, you can set caching rules up to ignore objects with certain strings in the URL eg ? or cgi-bin.. my logic is that if you get this right you can prevent the cache from storing objects which are nearly always dynamic and therefore increase the hit ratio as the objects on the drive end up being more static .. theres a similar section for non-cacheable mime-types. Theres also a section for cookie cheats to allow you to cache static objects within cookies. My question is has anyone spent time playing with these 3 fields and does anyone have some values which they have found to improve hit ratio that I can add to my (almost empty) config? On a different thread, if you add a netcache license how do you delete it, "DISABLE" doesnt work, the only way I found was to reinstall the OS on the filer. On -another- thread, I'm running WCCP and of course all web requests show the IP on the cache not the client itself.. is there a way to spoof the client's ip? I dont have a L4 switch btw, i run WCCP from a CAT6000 with MSFC. Thanks Steve -- Stephen J. Wilcox Internet Manager, Opal Telecom http://www.opaltelecom.co.uk/ Tel: 0161 222 2000 Fax: 0161 222 2008

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