On Wednesday 21 July 2004 12:52, Lee Kransen wrote:
The machine I need to copy will be totally idle,
except for normal system
activity, with no active users, connections or intentionally running
programs. The reason I thought to do it "live" is that the machine has no
CD or diskette drive and not even a USB port. I need to update the machine
every few days with the newest version of our system and then copy it to
several 10's of new machines.
If I understand correctly, it seems that the best option is for me to take
out the hard drive, put it in the DHCP/TFTP server and run the udp-sender
from there. Is that correct?
That's one possibility.
Other possibilities include:
1. Booting udpsender from network (DHCP) on that machine
2. or, Putting udpcast's kernel and initrd on the machine, and set up a
lilo (or grub) entry to boot that.
3. or, Do the udpcast "live" anyways, but from a low runlevel (1), with
root fs unmounted. As the FS is unmounted, the system is not really
live as far as udpcast is concerned...
If I decide to copy the disk to an image file and want
to zero-ize the
empty part of the drive, I understand that I can do this with dd. Could you
tell me the command?
On the live system, just create a huge file:
dd if=/dev/zero of=filler-file
After a while, the dd will error out with a "disk full" error.
Then, do a couple of syncs, and remove the file.
sync
sync
rm filler-file
If I want to use lzop or gzip to UDPCast a saved
image, do I have to save
the image in the appropriate compressed format before the transfer or does
UDPCast compress a saved image file on the fly during transfer?
Thanks a bunch,
Lee
Udpcast is able to compress the file on the fly (using the -p flag),
after all, that's how the bootable udpsender does it.
However, if you keep images on the server, it's best to precrompress
them on the server for two reasons:
1. Space saving (obvious...)
2. CPU time & disk read time saving (may lead to a faster transfer,
esp for those long stretches of zeroes)
Alain