toasters October 1999

toasters@lists.teaparty.net

81 participants
89 discussions

Re: General VIF questions
by Sam Cramer 26 Oct '99

26 Oct '99

cramer> Which interface takes over is specified using the "ifconfig" cramer> "partner" clause. Please note that vif partners MUST be specified cramer> using the interface name (e.g. vif0), not the interface address cramer> (e.g. 172.27.5.50). This was not clearly spelled out in the cramer> documentation, and it has tripped up a number of people. Brian> Hrm, I'm not sure what you mean by using the interface address Brian> instead of the IP address. If I just give the partner the VIF name, Brian> how does it know what IP to use? Is that stored on the mailbox Brian> disk? When a node takes over for its partner, it reads the partner's /etc/rc file. The ifconfig commands are processed specially: they are examined for a "partner" clause. If a partner clause is found, the interface named in the partner clause takes over for the interface specified in the ifconfig command. If no partner clause is found, the ifconfig command is ignored. Sam

3 2

General VIF questions
by Brian Tao 26 Oct '99

26 Oct '99

I have two clustered pairs of F740's arriving soon, so I will likely be able to discover answers to most of my questions, but I figured I'd do a bit of legwork and check my assumptions while I'm still in the planning stages. I'll consider just one of the clustered pairs: nfs1 and nfs2. Both will be equipped with two QFE NIC's. They will connect to a pair of switches with an uplink between them (i.e., passing traffic between two hosts on the same subnet, but on different switches). Logically, consider it a single switch. Each Netapp will be visible on two subnets: 10.35.8.0/25 and 10.35.8.128/25. NFS clients in one subnet will not be able to see filer interfaces in the other. I want to create two single-mode VIF's on each filer, using four interfaces on each filer: ====Active VIF_Subnet1====[e1a] [e2a]----Standby VIF_Subnet1--- ----Standby VIF_Subnet2---[e1b] [e2b]====Active VIF_Subnet2==== ..........Unused..........[e1c] [e2c]..........Unused.......... ..........Unused..........[e1d] [e2d]..........Unused.......... Trunks are created across NIC's. Ports on e1 are cabled to switch 1, and ports on e2 are cabled to switch 2. Since this is a single- mode trunk, the switches shouldn't care, right? During normal operation, switch 1 only sees an active port from e1a, and switch 2 only sees an active port from e2b (since e2a and e1b are effectively "downed" interfaces). While it would be a bit less confusing to say "all interfaces on e1 are active, and all interfaces on e2 are on standby", I prefer to have an active interface on each NIC at all times. This should lessen the chance of Murphy's Law kicking in on an early Sunday morning when e1 does fail, only to reveal that e2 had also died quietly sometime before then. I also want those same physical interfaces to assume their partner's VIF configuration in a failover situation (i.e., what the manual calls "shared interface takover"). So this is what I have, based on the 5.3.2 documentation: nfs1> vif create single VIF_Subnet1 e1a e2a nfs1> vif create single VIF_Subnet2 e2b e1b nfs1> vif favor e1a nfs1> vif favor e2b nfs1> ifconfig VIF_Subnet1 10.35.8.1 netmask 255.255.255.128 partner 10.35.8.2 nfs1> ifconfig VIF_Subnet2 10.35.8.129 netmask 255.255.255.128 partner 10.35.8.130 nfs2> vif create single VIF_Subnet1 e1a e2a nfs2> vif create single VIF_Subnet2 e2b e1b nfs2> vif favor e1a nfs2> vif favor e2b nfs2> ifconfig VIF_Subnet1 10.35.8.2 netmask 255.255.255.128 partner 10.35.8.1 nfs2> ifconfig VIF_Subnet2 10.35.8.130 netmask 255.255.255.128 partner 10.35.8.129 This configuration should be able to withstand failure of a filer NIC port, a filer NIC, an Ethernet cable, a switch port, an entire switch and an entire filer. Certain combinations of double failures are also non-lethal. Since this is all theoretical (to me, anyway), I have some questions. :) What characters are allowed in a VIF name? Is anyone using single-mode VIF's across two different switches like this? How does the Netapp know which VIF to use with the "vif favor" command? I assume this means a physical interface can only participate in one VIF at a time. Is it possible to prioritize interfaces in a VIF with more than two ports trunked together? Can VIF's be shared in a cluster? Can they share ports as above? Special mention is only made about super VIF's and clustering in the 5.3.2 SAG. Can super VIF's be created from single-mode VIF's, or only multi-mode VIF's? Again, the manual only mentions the latter, but does not prohibit the former. Can super VIF's themselves be multi-mode, or single-mode only? I'll post the results of my configuration testing once I have the hardware to play with. -- Brian Tao (BT300, taob(a)risc.org) "Though this be madness, yet there is method in't"

2 3

dlt8000 supported?
by Luke Mewburn 26 Oct '99

26 Oct '99

Hi All. Is the DLT-8000 supported in OnTap? I've been quoted a price on a DLT-8000 that is A$100 more than a DLT-7000, and when you're paying nearly A$8000 for a tape drive that seems like a good deal to me. Hopefully a DLT-8000 should act like a DLT-7000 given the same tape density SCSI control codes. I'd rather spend a little bit more money now for a drive with 12.5% more capacity, even if I have to wait for a newer release to get that extra density. Luke.

1 0

Re: mailboxes on toaster
by Jim McCoy 26 Oct '99

26 Oct '99

George Kahler wrote: >Q: Will mbx buy me performance on the toaster ? >Q: Is Maildir a better choice ? As in all difficult questions the answer is "it depends." People have given some very good reasons on why maildir is an appropriate choice given certain constraints, so let me mention a few things that favor mbox-format mail storage on a filer. We run several tens of millions of user mailboxes on a fleet of NetApps for YahooMail and so here are a few other points to consider that we have run into over the past few years: 1) Locking is a dead issue for mbox. There is no way to adequately provide file locking via the standard channels so you will probably need to deal with the locking issue as a seperate problem. We ended up writing our own mailbox locking system but you might be able to use procmail for as your local delivery agent and let that try to figure out the locking issues. Locking is important, but do not fixate on this issue and ignore all of the other points that you need to consider. 2) Inode cost should not be overlooked. You get at most 32 million per volume, so if you have a lot of mailboxes a maildir or other one-file-per-message scheme will chew through your inode limit long before you run out of disk space on that volume. To compensate you will have to make smaller volumes. This is not necessarily a bad thing but each volume will be at least one additional raid group and on additional wasted parity disk (increased cost/mailbox). The larger number of inodes in use will also increase the time it takes your filer to complete a wack run and will have an impact on backup speeds. 3) Mail messages are small. There is a simple bimodal distribution to internet mail message size, they are either very small or very large. The small files will cost you storage bits because the filer uses 4K to store each 500byte message. The small file sizes will also keep your read chain length very low so we will get significantly decreased performance on your backups of this volume and you will notice some decreased read performance on the filer in general. jim

4 7

Re: mailboxes on toaster
by jrg＠gbnet.net 25 Oct '99

25 Oct '99

On Mon 25 Oct, 1999, "Luke Gain" <luke(a)erinet.com> wrote: >one thing I found to be a big win is to make sure all the filenames are 12 >characters >or less. evidentally, the netapp only caches names that are <12 characters >long. >so if the name is >12 characters it takes a disk-op to read the directory. I expect it'll interact with your operating system, too (or this may actually be the sole cause.) Eg. Solaris puts namei lookup results in the DNLC (directory name lookup cache) only if they're under a certain size (vmstat -s shows these as "toolong"), so if you've got long directory paths you would expect to see a problem not dissimilar to the one you describe. On the subject of the DNLC, beware of making it too big as that has been bad for performance too. ("classic" BSD sets this "toolong" constant NCHNAMLEN at 31. FreeBSD-current seems to have done away with it.) James.

1 0

Re: General VIF questions
by Sam Cramer 25 Oct '99

25 Oct '99

>>>>> "Brian" == Brian Tao <taob(a)risc.org> writes: Brian> Since this is all theoretical (to me, anyway), I have some Brian> questions. :) Brian> What characters are allowed in a VIF name? A vif name may be composed of any non-blank characters and should begin with a letter. The name may be at most 15 characters long. Brian> Is anyone using single-mode VIF's across two different switches like Brian> this? Yes. We have a number of customers who use single-mode vifs composed of links attached to two different switches to protect against switch failure. Brian> How does the Netapp know which VIF to use with the "vif favor" Brian> command? I assume this means a physical interface can only Brian> participate in one VIF at a time. Yes, an interface (physical or virtual) can only participate in one vif at a time. Brian> Is it possible to prioritize interfaces in a VIF with more than two Brian> ports trunked together? If you're asking "can you favor a link in a single-mode VIF composed of more than 2 links", the answer is yes. If you're asking "can I specify secondary, tertiary (and so on) priority for links in a single-mode VIF", the answer is no: only one link may be favored; all others are equally "non-favored". Brian> Can VIF's be shared in a cluster? Can they share ports as above? Brian> Special mention is only made about super VIF's and clustering in the Brian> 5.3.2 SAG. vifs can't be shared across a cluster. However, the surviving node may takeover for a vif just like it may takeover for other interfaces. Which interface takes over is specified using the "ifconfig" "partner" clause. Please note that vif partners MUST be specified using the interface name (e.g. vif0), not the interface address (e.g. 172.27.5.50). This was not clearly spelled out in the documentation, and it has tripped up a number of people. Brian> Can super VIF's be created from single-mode VIF's, or only Brian> multi-mode VIF's? Again, the manual only mentions the latter, but Brian> does not prohibit the former. Super vifs may be created from single-mode vifs or multi-mode vifs. I'm not sure why you would want to create a super-vif from single-mode vifs, though. Brian> Can super VIF's themselves be multi-mode, or single-mode only? Single mode only. Sam

2 1

We Need Homeworkers!!
by mowooco＠mindspring.com 25 Oct '99

25 Oct '99

Dear Future Associate, You Can Work At Home & Set Your Own Hours. Start earning Big Money in a short time NO Newspaper Advertising! Your job will be to stuff and mail envelopes for our company. You will receive $.25 for each and every envelope you stuff and mail out. Just follow our simple instructions and you will be making money as easy as 1 2 3 For example stuff and mail 200 envelopes and you will receive $50.00. Stuff and mail 1000 and you will receive $250.00. Stuff and mail 2000 and you will receive $500.00 and more Never before has there been an easier way to make money from home! Our Company's Home Mailing Program is designed for people with little or no experience and provides simple, step by step instructions. There is no prior experience or special skills necessary on your part, Just stuffing envelopes. We need the help of honest and reliable home workers like you. Because we are overloaded with work and have more than our staff can handle. We have now expanded our mailing program and are expecting to reach millions more with our offers throughout the US and Canada. Our system of stuffing and mailing envelopes is very simple and easy to do! You will not be required to buy envelopes or postage stamps. We will gladly furnish all circulars at no cost to you. We assure you that as a participant in our program you will never have to mail anything objective or offensive. There are no quotas to meet, and there no contracts to sign. You can work as much, or as little as you want. Payment for each envelope you send out is Guaranteed! Here is what you will receive when you get your first Package. Inside you will find 100 envelopes, 100 labels and 100 sales letters ready to stuff and mail As soon as you are done with stuffing and mailing these first letters, your payment will arrive shortly, thereafter. All you have to do is to order more free supplies and stuff and mail more envelopes to make more money. Our sales literature which you will be stuffing and mailing will contain information outlining our highly informative manuals that we are advertising nationwide. As a free gift you will receive a special manual valued at $24.95, absolutely free, just for joining our Home Mailers Program. Plus you will get your own special code number, so that we will know how much you are to get paid. And to make re-ordering of more envelopes, that our company supplies very simple for you. We are giving you this free bonus because we want you to be confident in our company and to ensure that we will be doing business with you for a long time. Benefits Of This Job: 1. You do not have to quit your present job, to earn more money at home 2. You can make between $2,500 to $4,500 a month depending on the amount of time you are willing to spend stuffing and mailing envelopes 3. This is a great opportunity for the students, mothers, disabled persons or those who are home bodies. To secure your position and to show us that you are serious about earning extra income at home we require a one-time registration fee of $35.00. This fee covers the cost of your initial start up package, which includes 100 envelopes, 100 labels and 100 sales letters and a manual, your registration fee will be refunded back to you shortly thereafter. Money Back Guarantee! We guarantee that as soon as you stuff and mail your first 300 envelopes You will be paid $75.00 and your registration fee will be refunded. Many of you wonder why it is necessary to pay a deposit to get a job. It is because we are looking for people that seriously want to work from home. * If 3.000 people told us they wanted to start working from home and we sent out 3.000 packages free to every one. And then half of the people decided not to work, this would be a potential loss of more than $60,000 in supply's and shipping that we have sent out to people that don't want to work We have instituted this policy to make sure that you really want to work and at least finish your first package. To Get Started Today Please Enclose Your Registration Fee of $35 Check,Money Order and fill out the application below and mail to: MOHW Co 11054 Ventura Blvd PMB #126 Studio City, CA 91604 Name_____________________________________________________ Address___________________________________________________ City____________________________________ State______________ Zip Code________________ Telephone Number(s)_________________________________________ E-mail Address______________________________________________ For all orders, please allow seven (7) days for delivery and up to 10 days. Money Orders will result in faster shipping of your package.

1 0

a recent caching company bugid ticket info
by Colin Johnston 22 Oct '99

22 Oct '99

For a good Friday night read :) Title Some customers would like our apppliances to be more groovy Problem It was noted by a customer that the design of our caches wasn't groovy enough. They would like the caches to glow and blink. Stackability and the ability to use the server as a die in Dungeons & Dragons games was also mentioned. Workaround You can currently customize the "groovability" of the cache with third party add ons from http://www.legomindstorms.com/ . Your imagination and pocketbook are your only limitations. Colin Johnston SA PSINET UK

1 0

Source for a non-root quota program?
by James Deibele 21 Oct '99

21 Oct '99

We would like our support people to be able to do real-time quota verification for customers, meaning that support person A is able to tell customer B that they're using 9.688MB of their 10MB quota. Having gone that far, we would actually really like to have a web page for our non-UNIX customers to check their usage. We can't use the normal /usr/ucb/quota program that comes with Solaris because it requires you to run as root to check someone else's usage. If we try to use suexec or cgiwrap to run programs as the user, we will have to make a copy of the script for every customer. This is sub-optimal. So what we want is a program (Solaris preferred, Linux or NT might be acceptable) that will allow any user to check the total usage for any other user. Seems like this would have come up with other people using NetApps extensively. Does NetApp have an official program that does this? Possibly a program that someone has contributed?

7 7

Re: Source for a non-root quota program?
by dnb＠ccs.neu.edu 20 Oct '99

20 Oct '99

I've not tried this as non-root, but there is a Perl module which can do all sort of good quota stuff for you. It can be found at: http://search.cpan.org/search?module=Quota Peace, dNb

1 0

← Newer
1
2
3
4
5
6
7
8
9
Older →

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 October 1999