'Cloning' a linux system

'Cloning' a linux system

Post by k.. » Wed, 09 Nov 1994 04:06:15



Many of the CS majors here at *ia Tech use decstations, running
Ultrix.  Whenever an update of system software is released, or some
security hole is discovered, or a really nifty tool (like Mosaic) is
compiled for their platform, they can get their decstations 'recloned'
at a computer lab.  I don't know a lot about the process, but it looks
like they just hook a SCSI cable between the decstation and run some
sort of search, compare, and destroy or copy program.

Cloning is supposedly a great headache-preventer for the support
people responsible for the CS machines.  

I think that something like this would be useful for Linux to have.
If we want to get huge institutions more interested in Linux, it would
be good to make it easier for them to make their own
tweaked-to-perfection reference setup of Linux that could then be
cloned out without too much difficulty.

Any thoughts on how this could be implemented in a useful manner?

-Ken Bateman

 
 
 

'Cloning' a linux system

Post by Andre Apr » Wed, 09 Nov 1994 16:56:20


: Many of the CS majors here at *ia Tech use decstations, running
: Ultrix.  Whenever an update of system software is released, or some
: security hole is discovered, or a really nifty tool (like Mosaic) is
: compiled for their platform, they can get their decstations 'recloned'
: at a computer lab.  I don't know a lot about the process, but it looks
: like they just hook a SCSI cable between the decstation and run some
: sort of search, compare, and destroy or copy program.

: Cloning is supposedly a great headache-preventer for the support
: people responsible for the CS machines.  

: I think that something like this would be useful for Linux to have.
: If we want to get huge institutions more interested in Linux, it would
: be good to make it easier for them to make their own
: tweaked-to-perfection reference setup of Linux that could then be
: cloned out without too much difficulty.

: Any thoughts on how this could be implemented in a useful manner?

: -Ken Bateman

Get 'rdist'. It does the job for us.

Andr.

 
 
 

'Cloning' a linux system

Post by Brad Matthew Garc » Sat, 12 Nov 1994 06:55:50


|> > Many of the CS majors here at *ia Tech use decstations, running
|> > Ultrix.  Whenever an update of system software is released, or some
|> > security hole is discovered, or a really nifty tool (like Mosaic) is
|> > compiled for their platform, they can get their decstations 'recloned'
|> > at a computer lab.  I don't know a lot about the process, but it looks
|> > like they just hook a SCSI cable between the decstation and run some
|> > sort of search, compare, and destroy or copy program.
|>
|> > Cloning is supposedly a great headache-preventer for the support
|> > people responsible for the CS machines.  
|>
|> > I think that something like this would be useful for Linux to have.
|> > If we want to get huge institutions more interested in Linux, it would
|> > be good to make it easier for them to make their own
|> > tweaked-to-perfection reference setup of Linux that could then be
|> > cloned out without too much difficulty.
|>
|> > Any thoughts on how this could be implemented in a useful manner?

Penn State University had a nifty way of keeping all their networked
PC's up to date with software.  You may want to contact someone at PSU
for more info.

Whenever one of these networked PC's is rebooted, its entire hard drive
is checked and updated from a server.  This way people can*around
with the hard drive all they want to (including putting DooM on it  :^)
and all you'd have to do to get it back to working order would be to
reboot it.  Maybe something like this would be even better?  
(Don't have to carry your computer to a lab).
--
Brad M. Garcia         "If the auto industry were like the computer industry,
   ____/  _ _ /  /   /  a car would now cost $50, would get 500 mpg, and at a
  /      /_/ /  /   /   random time would explode, killing all passengers."
_____/ _/  _/ _____/                                     -unknown source

 
 
 

'Cloning' a linux system

Post by Francisco A.Garc » Sat, 12 Nov 1994 05:14:46



> Many of the CS majors here at *ia Tech use decstations, running
> Ultrix.  Whenever an update of system software is released, or some
> security hole is discovered, or a really nifty tool (like Mosaic) is
> compiled for their platform, they can get their decstations 'recloned'
> at a computer lab.  I don't know a lot about the process, but it looks
> like they just hook a SCSI cable between the decstation and run some
> sort of search, compare, and destroy or copy program.
> Cloning is supposedly a great headache-preventer for the support
> people responsible for the CS machines.  
> I think that something like this would be useful for Linux to have.
> If we want to get huge institutions more interested in Linux, it would
> be good to make it easier for them to make their own
> tweaked-to-perfection reference setup of Linux that could then be
> cloned out without too much difficulty.
> Any thoughts on how this could be implemented in a useful manner?

        One * way would be to run some type of diff between the old
& new versions & changing that. ie. copying a new kernel remotely etc.
Or just pass around a tape backup ;)

Quote:> -Ken Bateman

--
=-----------------------------------------------------------------------------=
= Computer Science Major.                   Ball State University, Muncie IN. =

=-----------------------------------------------------------------------------=
 
 
 

'Cloning' a linux system

Post by Gary Houst » Sun, 13 Nov 1994 12:03:17





> > I think that something like this would be useful for Linux to have.
> > If we want to get huge institutions more interested in Linux, it would
> > be good to make it easier for them to make their own
> > tweaked-to-perfection reference setup of Linux that could then be
> > cloned out without too much difficulty.

> > Any thoughts on how this could be implemented in a useful manner?

>    One * way would be to run some type of diff between the old
> & new versions & changing that. ie. copying a new kernel remotely etc.
> Or just pass around a tape backup ;)

You could try the "track" system, where a bunch of machines can
check automatically for updates from a server, e.g., when they boot
or at night from cron.  It can also run programs, such as lilo after
updating the kernel.  It takes some work to get it running however.
 
 
 

'Cloning' a linux system

Post by Steve Clar » Tue, 15 Nov 1994 00:37:23



> Cloning is supposedly a great headache-preventer for the support
> people responsible for the CS machines.  
> Any thoughts on how this could be implemented in a useful manner?

An extremely * way (but very effective) that only works if all machine
hardware is identical is to use dd on the raw install disk, make a bakup copy
elsewhere on the net, then use dd to pull it back onto the new machine.

Note that hostname/ip addresses will need re-configuring afterwards!

S.

 
 
 

'Cloning' a linux system

Post by Mark Aitchison - Physics and Astronomy Computologi » Thu, 17 Nov 1994 14:03:49


I've successfully cloned hard disks containing DOS and Linux partitions using a
DOS program reading tracks under DOS over the network - the result assumes no
bad blocks, and requires any individual files (e.g. its own IP number) to then
be editted. But is darned fast, and you know that there can be no virus or
whatever left after it!  If anybody is interested in the programs at each end
(they are darned simple!) let me know.

Mark Aitchison.

 
 
 

'Cloning' a linux system

Post by Chris Ulri » Fri, 18 Nov 1994 18:12:03




Quote:>I've successfully cloned hard disks containing DOS and Linux partitions using a
>DOS program reading tracks under DOS over the network - the result assumes no
>bad blocks, and requires any individual files (e.g. its own IP number) to then
>be editted. But is darned fast, and you know that there can be no virus or
>whatever left after it!  If anybody is interested in the programs at each end
>(they are darned simple!) let me know.

the program dd should do the same thing.  It is also fast, probably faster.
I guess cat should also do the same thing, clone an entire disk or partition.
Again, bad blocks and such will bit, but who cares with modern disks.
(dd if=/dev/sda of=/dev/sdb)  Just dont get the order wrong, otherwise you
will wind up with a very blank disk and you will probably be very unhappy.
chris
 
 
 

'Cloning' a linux system

Post by Alan C » Sat, 19 Nov 1994 21:41:22




>> Any thoughts on how this could be implemented in a useful manner?
>    One * way would be to run some type of diff between the old
>& new versions & changing that. ie. copying a new kernel remotely etc.
>Or just pass around a tape backup ;)

man rdist

Alan
--
  ..-----------,,----------------------------,,----------------------------,,

 ``----------'`----------------------------'`----------------------------''

 
 
 

'Cloning' a linux system

Post by Craig I. Hag » Sun, 20 Nov 1994 07:19:03



> Many of the CS majors here at *ia Tech use decstations, running
> Ultrix.  Whenever an update of system software is released, or some
> security hole is discovered, or a really nifty tool (like Mosaic) is
> compiled for their platform, they can get their decstations 'recloned'
> at a computer lab.  I don't know a lot about the process, but it looks
> like they just hook a SCSI cable between the decstation and run some
> sort of search, compare, and destroy or copy program.
> Cloning is supposedly a great headache-preventer for the support
> people responsible for the CS machines.  
> I think that something like this would be useful for Linux to have.
> If we want to get huge institutions more interested in Linux, it would
> be good to make it easier for them to make their own
> tweaked-to-perfection reference setup of Linux that could then be
> cloned out without too much difficulty.
> Any thoughts on how this could be implemented in a useful manner?
> -Ken Bateman

look into sup utility from CMU. It is on sunsite somewhere.

-- craig

 
 
 

'Cloning' a linux system

Post by Van Zan » Wed, 23 Nov 1994 02:52:03




>> Cloning is supposedly a great headache-preventer for the support
>> people responsible for the CS machines.  
>> Any thoughts on how this could be implemented in a useful manner?

...
>Note that hostname/ip addresses will need re-configuring afterwards!

Note that according to the file system standard, /usr can be shared
among several machines.  Anything that is machine specific is somewhere
else (like /etc or /var) with symbolic links in /usr if necessary.  As
long as your changes are in /usr, you ought to be able to copy the
whole directory over, any way you like.  Or NSF mount /usr to begin with,
and only maintain one copy (probably slow, though).

                             - Jim Van Zandt

 
 
 

'Cloning' a linux system

Post by Donald Beck » Fri, 25 Nov 1994 06:07:13





>>I've successfully cloned hard disks containing DOS and Linux partitions using a
>>DOS program reading tracks under DOS over the network - the result assumes no
>>bad blocks, and requires any individual files (e.g. its own IP number) to then
>>be editted. But is darned fast, and you know that there can be no virus or
>>whatever left after it!  If anybody is interested in the programs at each end
>>(they are darned simple!) let me know.

>the program dd should do the same thing.  It is also fast, probably faster.
>I guess cat should also do the same thing, clone an entire disk or partition.
>Again, bad blocks and such will bit, but who cares with modern disks.
>(dd if=/dev/sda of=/dev/sdb)  Just dont get the order wrong, otherwise you
>will wind up with a very blank disk and you will probably be very unhappy.
>chris

I had to duplicate a lot of 540M IDE disks for the Beowulf cluster, so I got
a chance to experiment.  Under Linux there is essentially no performance
difference between the following commands:
        cp /dev/hda /dev/hdb
        cat /dev/hda > /dev/hdb
        dd if=/dev/hda of=/dev/hdb
I did most subsequent tests with the 'cp' version, since they ugly 'dd'
syntax reminds me of PIP.

This kind of copy is only safe with IDE disks that are configured for the
same virtual C/H/S (cylinder/...) parameters.  It works because the IDE
interface presents a perfect virtual disk.  Bad blocks are transparently
mapped to blocks reserved for this purpose.  If your non-IDE (e.g. ST506,
MFM, SCSI) disk interface doesn't hide bad blocks it won't work.  If the
C/H/S values are different your partition table and filesystems info will be
inconsistent.

Notes
The disks were Maxtor 7546AT 521MB IDE disks with 256K cache.
The controllers were generic $25 ones -- DTC 2278VL w/ a DTC805 IDE chip.
The baseline 1.0.* kernel took about 30 minutes to copy 540M between disks.
The biggest improvents came from changing to a 1.1.* kernel with Mark Lord's
changes, which dropped the time down to about 9 minutes.
Further improvements came from Mark Lord 'hdparm' program to turn on
multiple sector reads and writes.  This dropped the copy time down to about
6 minutes, an average rate of 3MB/sec.

--

USRA-CESDIS, Center of Excellence in Space Data and Information Sciences.
Code 930.5, Goddard Space Flight Center,  Greenbelt, MD.  20771
301-286-0882         http://cesdis.gsfc.nasa.gov/pub/people/becker/whoiam.html

 
 
 

1. What's the best linux software for cloning audio CD's?

I need to copy some audio CD's to take out to live band gigs for break
music so I don't scratch up the masters. What's the the best software
for cloning audio CD's? I have XCDRoast, but I get nothing but digital
static when I try to copy audio CD's. I have tried various settings
mentioned in the help file to no avail. Is there some other simple
software package that will just make an accurate copy of an audio CD?

2. 3.2v4: any 5.1/4 uod429a disk?

3. Users having probs logging into Linux system, get '/home/...' doesn't exist.

4. Please help w/ emacs 19.31 core dump

5. copying / cloning a system disk, can't find an exact answer.

6. Modem suggestions please?

7. ''Charlie Root'' System Monitor - TURN OFF

8. Electric-Fence slooooow in 2.1.[45]X kernels: Why

9. what is 'protection fault' and 'system calls' ?

10. Problem running 'waisq' in 'chroot' file system

11. system doesn't recognize ne2000 clone

12. Implement '<'or '>' system commands in C

13. HELP: system won't recognize PAS16 clone