Re: Mass install / Autoinstall (Was: Re: Debian vs Red Hat??? I need info.)
Hello All There is definitately some scope for development in this area. Debian is one of the best distro's to maintain but it is one of the worst to install. These advantages and disadvantages are multiplied when you have many machines to maintain. On 17 May 00, at 21:55, Karl M. Hegbloom wrote: You can make a copy of the system like this... it will create a `cpio' archive... substitute `ustar' for `crc' to make a `tar' compatible archive. RTFM's... you're on your own. 88 #!/bin/bash find / -print0 | grep --invert-match --extended-regexp --null-data --file=/root/make-tarball.exclude-patterns | cpio --create --format=crc --null --reset-access-time --block-size=10 | gzip --best /tmp/system-snapshot_$(date +%Y.%m.%d).cpio.crc.gz 88 I tried this to create a custom base2_2.tzg with reasonable results. First problem is that we need a tar file and not a cpio one. Cpio's tar format does not support block devices so the whole /dev/ directory gets broken. Then I tried ustar. This worked better but still has some limitation on file name length. A few files in /var/state/apt/lists/* were too long - not a major trainsmash. I wasted a few hours trying tar instead of cpio. It seems not to be able to backup a directory, without backing up the contents of that directory, this is a problem with things like /var/cache/apt/archives. Maybe a real find/grep/tar guru could get it right but I went back to Karl's script :-) I still have some bugs. After the base install lilo would not run (something broken with vmlinuz softlink). Then when the new system is rebooted it went into a loop asking about shadow passwords etc. I eventually replaced the /etc/inittab. Bug squashing is a slow process ... a full test cycle requires a backup and a new installation. This seems a viable method of setting up a mass install system. After I got things going I used Midnight Commander to do some global searches and replaces in /etc to sort out things like domain names and ppp accounts etc and then I had a system ready to run with exim, squid, dns ,ppp, diald, mgetty, calamaris, dhcp, apache, ftp, ipchains, samba, uucp, fetchmail etc all working! Best of all it is a fully compliant Debian system, so apt-get update| apt-get upgrade also works! Next step may be to modify the dinstall program. Question: Is'nt there a deb package with scripts for creating boot disks? I feel I should not be reinventing the wheel. Another question: Which list should we be discussing this? Karl's original messages was sent to a whole bunch of lists? My modified scripts are as follows (mind the line wrapping): #! /bin/bash find / -print0 | grep --invert-match --extended-regexp --null-data -- file=/root/config/exclude-pattern | cpio --create --format=ustar --null --reset-access-time --block- size=10 | gzip --best /tmp/base2_2-$(date +%Y.%m.%d).tgz ^/proc/.* ^/tmp/.* /lost+found ^/boot/lost+found ^/var/cache/apache/.* ^/var/cache/apt/.*\.deb ^/var/log/.*\.log ^/var/log/\(amanda\|apache\|gdm\|ksymoops\|mailman\|news\|sendfil e\|wu-ftpd\)/.* ^/var/log/\(syslog\|smb\|nmb\|messages\|mail\|lpr\|debug\|dmesg\).* ^/var/lock/.* ^/var/run/.*\.pid ^/var/run/\(ndc\|utmp\) ^/var/samba/.* ^/var/spool/squid/.*/.*/.* \.bash_history \.gnome-errors .*~ /\.saves-.* /\.#.* /\.netscape/cache/.* ^/etc/modules ^/etc/hostname ^/etc/hosts ^/etc/networks ^/etc/resolv.conf ^/etc/modutils/ ^/etc/apm/event.d/pcmcia ^/etc/init.d/pcmcia ^/etc/pcmcia/ ^/etc/network/interfaces ^/tmp/ Ian Forbes - Ian Forbes ZSD http://www.zsd.co.za Office: +27 +21 683-1388 Fax: +27 +21 64-1106 Snail Mail: P.O. Box 46827, Glosderry, 7702, South Africa -
RE: Mass install / Autoinstall (Was: Re: Debian vs Red Hat??? I need info.)
Long time ago in the past '[EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]' wrote copy everything from the master drive to the copy, then run the appropriate Lilo command to make that copy bootable. You can then mount it in another machine and it's ready to go. You have to filter some things out when you copy. See below. You can't do that, I've tried it before. Lilo can't be installed on any secondary disk. Don't ask me why because I don't know. There's a HOWTO about it. -- Hi to * ! ( First of all, sorry for my english :) IMHO LILO can be installed on secondary disks ( and on the first sector of a bootable partition also ). Let me explain myself: I have 2 HDD at home, one ( /dev/hda ) with M$ Win 95 WinNT 'PlayStation' ( Job reasons :o( ) and the other ( /dev/hdc ) with Debian Hamm ( 'So old' I know ;oD I'm try to upgrade but... unfortunately 'time is no friend of mine' :( ). I use (Commercial SW) V Communications' System Commander ( http://www.v-com.com/ ) to boot my machine with Linux and (gasp!) with M$ Win XX, and it works fine for me! LILO was installed on MBR of the secondary disk ( /dev/hdc ), it shows an alert 'LILO is not installed on primary disk' or something like that... but it runs ;) Try to add these lines ( don't remember exactly, I'm at the office right now ) to your /etc/lilo.conf: ... ... boot=/dev/hdc # my 2nd HDD ... ... image=/boot/vmlinuz label=linux root=/dev/hdc1 # or whatever your partition is ;o) read-only I'm no Linux guru, so please correct me if I've misunderstood you or I'm wrong with anything. Regards from Spain, Santy #;oD -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Mass install / Autoinstall (Was: Re: Debian vs Red Hat??? I need info.)
Jeremy == Jeremy Hansen [EMAIL PROTECTED] writes: Jeremy Autoinstall (Red Hat's kickstart) Jeremy This is also something fairly important. We need this as we do a Jeremy lot of mass installs. The best way to do that that I've found so far is to set up a box with two removable hard drive racks, install and _configure_ everything on one drive, then use `cfdisk', `mkswap', and `mke2fs' to partition and format the second drive. Use `cpio' from a script to copy everything from the master drive to the copy, then run the appropriate Lilo command to make that copy bootable. You can then mount it in another machine and it's ready to go. You have to filter some things out when you copy. See below. Another way to do it would be to create a tar archive, useing find | grep -v -f exclude-patterns | cpio, name it `base2_2.tgz' and put it in place on an intranet web server where you can point the Debian installer's netfetch... Then you can install several machines at once over the LAN... in theory. This is just a starter... I have not done this much yet myself, since I don't have extra hardware to work with and really need to spend my time on reading and studies. I have done it from drive to drive using `cpio' to install the filesystem snapshot, but have not done it by naming a tar format archive as base and using the debian-boot installer. It might just work. NFS mounting the server directory where the `cpio' or `tar' archive sits might work fine also. You could burn a bootable CD with the archive on it, and on the bootable's root.bin, have `sfdisk' etc. and a script that automaticly partitions, formats, and installs the archive. It might be simpler to try the netfetch/dbootstrap approach though. You can make a copy of the system like this... it will create a `cpio' archive... substitute `ustar' for `crc' to make a `tar' compatible archive. RTFM's... you're on your own. 88 #!/bin/bash find / -print0 | grep --invert-match --extended-regexp --null-data --file=/root/make-tarball.exclude-patterns | cpio --create --format=crc --null --reset-access-time --block-size=10 | gzip --best /tmp/system-snapshot_$(date +%Y.%m.%d).cpio.crc.gz 88 You may need to tweak this some. (NO WARRANTEE) make-tarball.exclude-patterns 88 ^/proc/.* ^/tmp/.* ^/lost+found ^/boot/lost+found ^/var/cache/apache/.* ^/var/cache/apt/.*\.deb ^/var/log/.*\.log ^/var/log/\(amanda\|apache\|gdm\|ksymoops\|mailman\|news\|sendfile\|wu-ftpd\)/.* ^/var/log/\(syslog\|smb\|nmb\|messages\|mail\|lpr\|debug\|dmesg\).* ^/var/lock/\.LCK.* ^/var/run/.*\.pid ^/var/run/\(ndc\|utmp\) ^/var/samba/.* \.bash_history \.gnome-errors .*~ /\.saves-.* /\.#.* /\.netscape/cache/.* -- Those who do not study Lisp are doomed to reimplement it - Poorly. A few months in the laboratory often saves several hours at the library. mailto:[EMAIL PROTECTED] (Karl M. Hegbloom)
Mass install / Autoinstall (Was: Re: Debian vs Red Hat??? I need info.)
[...] KMH The best way to do that that I've found so far is to set up KMH a box with two removable hard drive racks, install and KMH _configure_ everything on one drive, then use `cfdisk', KMH `mkswap', and `mke2fs' to partition and format the second KMH drive. [...] I do a possibly non-kosher thing similar to the above. I tar everything up once it is set up and stick the tar file[s] into a SCSI drive. I have a box that boots from this SCSI drive and has IDE drawers and a kernel with IDE support built as modules. I then hot-swap IDE drives, sfdisk, mke2fs, mount and un-tar without bringing down the machine. Insmoding the ide modules after switching the drives on and rmmoding before removing them seems to work fine. Never lost a drive yet, but the largest drives I worked with under this scheme were 4.3G. With the newer/larger drives, you'd probably need to make sure LILO and the BIOS agree on a geometry for the drive to be actually bootable (dunno the incantation for that yet!). cheers, BM
Re: Mass install / Autoinstall (Was: Re: Debian vs Red Hat??? I need info.)
Most of the answers I've been getting on this subject seem like total hacks, which may work but really are tricks to doing this. I was really looking for something within debian that's built to do kickstart type installations. Although what you suggest may work, it leave little flexibility between machines and also takes a lot more work then I was hoping to do. For example, I have 20 machines at a co location I need to go install. Right now with Red Hat I can take my laptop, slap a floppy in each machine, turn 'em on, 5 minutes later I have 20 fully configured machines ready to rock. Also if I use DHCP and place my kick start config file on the server, I could literally have 20 different configurations for each machine and never have to touch a key. This is a part of Red Hat, no tricks have to be done, all you need is a proper ks.cfg file and a central place where the distro comes from, usually over nfs for convenience. YOu can't beat that when doing large installations. To do what I need to do in Debian seems that it would take a very long time, even hours, which is not fun if you've ever spent time at a co location. It seems a lot of Debian users are developers and in this case I'm sure Debian is perfect, but Red Hat's kickstart allows me to see my wife at night (not really, but you know what I mean). -jeremy Jeremy == Jeremy Hansen [EMAIL PROTECTED] writes: Jeremy Autoinstall (Red Hat's kickstart) Jeremy This is also something fairly important. We need this as we do a Jeremy lot of mass installs. The best way to do that that I've found so far is to set up a box with two removable hard drive racks, install and _configure_ everything on one drive, then use `cfdisk', `mkswap', and `mke2fs' to partition and format the second drive. Use `cpio' from a script to copy everything from the master drive to the copy, then run the appropriate Lilo command to make that copy bootable. You can then mount it in another machine and it's ready to go. You have to filter some things out when you copy. See below. Another way to do it would be to create a tar archive, useing find | grep -v -f exclude-patterns | cpio, name it `base2_2.tgz' and put it in place on an intranet web server where you can point the Debian installer's netfetch... Then you can install several machines at once over the LAN... in theory. This is just a starter... I have not done this much yet myself, since I don't have extra hardware to work with and really need to spend my time on reading and studies. I have done it from drive to drive using `cpio' to install the filesystem snapshot, but have not done it by naming a tar format archive as base and using the debian-boot installer. It might just work. NFS mounting the server directory where the `cpio' or `tar' archive sits might work fine also. You could burn a bootable CD with the archive on it, and on the bootable's root.bin, have `sfdisk' etc. and a script that automaticly partitions, formats, and installs the archive. It might be simpler to try the netfetch/dbootstrap approach though. You can make a copy of the system like this... it will create a `cpio' archive... substitute `ustar' for `crc' to make a `tar' compatible archive. RTFM's... you're on your own. 88 #!/bin/bash find / -print0 | grep --invert-match --extended-regexp --null-data --file=/root/make-tarball.exclude-patterns | cpio --create --format=crc --null --reset-access-time --block-size=10 | gzip --best /tmp/system-snapshot_$(date +%Y.%m.%d).cpio.crc.gz 88 You may need to tweak this some. (NO WARRANTEE) make-tarball.exclude-patterns 88 ^/proc/.* ^/tmp/.* ^/lost+found ^/boot/lost+found ^/var/cache/apache/.* ^/var/cache/apt/.*\.deb ^/var/log/.*\.log ^/var/log/\(amanda\|apache\|gdm\|ksymoops\|mailman\|news\|sendfile\|wu-ftpd\)/.* ^/var/log/\(syslog\|smb\|nmb\|messages\|mail\|lpr\|debug\|dmesg\).* ^/var/lock/\.LCK.* ^/var/run/.*\.pid ^/var/run/\(ndc\|utmp\) ^/var/samba/.* \.bash_history \.gnome-errors .*~ /\.saves-.* /\.#.* /\.netscape/cache/.* -- http://www.xxedgexx.com | [EMAIL PROTECTED] -
Re: Mass install / Autoinstall (Was: Re: Debian vs Red Hat??? I need info.)
Are you aware of this? http://www.informatik.uni-koeln.de/fai/ -- Mike On 2000-05-18 at 13:55 -0400, Jeremy Hansen wrote: It seems a lot of Debian users are developers and in this case I'm sure Debian is perfect, but Red Hat's kickstart allows me to see my wife at night (not really, but you know what I mean).
Re: Mass install / Autoinstall (Was: Re: Debian vs Red Hat??? I need info.)
I would agree most of the proposed solutions are quick hacks. The fact is, we won't be natively supporting bulk installation until Woody. And even that is in question. As I understand it, the proposed Woody install system is debconf based; moreover, debconf can have different backends for receiving configuration info, for instance, an LDAP backend, or a backend that munges an XML file from a web server. Yes, vapor vapor vapor but that's the right way to do it if you ask me. Hopefully debconf will be _de rigeur_ for any package requiring configuration info at pkg install time in Woody, so what we would have is really a general solution rather than just a partial or hack solution. -- .Adam Di [EMAIL PROTECTED]URL:http://www.onShore.com/
Re: Mass install / Autoinstall (Was: Re: Debian vs Red Hat??? I need info.)
Agreed that this seems technically sound, but it would be really nice to have this Real Soon Now. I think it might be reasonably possible to backport this from Woody into Potato fairly soon after the release of Potato. The fact is that an automatic installation system will be really hard to test on the unstable tree. I am not proposing that something like this should really be called stable, but if it could be made compatible with the stable distribution (then Potato) that would be very helpful. -- Mike On 2000-05-18 at 19:32 -0400, Adam Di Carlo wrote: The fact is, we won't be natively supporting bulk installation until Woody. And even that is in question. As I understand it, the proposed Woody install system is debconf based; moreover, debconf can have different backends for receiving configuration info, for instance, an LDAP backend, or a backend that munges an XML file from a web server.
Re: Mass install / Autoinstall (Was: Re: Debian vs Red Hat??? I need info.)
On Thu, May 18, 2000 at 01:55:37PM -0400, Jeremy Hansen wrote: Most of the answers I've been getting on this subject seem like total hacks, which may work but really are tricks to doing this. I was really looking for something within debian that's built to do kickstart type installations. huh? what do you think kickstart is? it's the same kind of total hack - the difference is that you have to do it RedHat's way whether you like it or not, and it pretends to be easy enough to use that you don't need to know what you're doing to run it. personally, i think that anyone who needs to mass-build machines *SHOULD* know exactly what they are doing. i wouldn't trust any machine built by someone who needed such point-and-click tools. Although what you suggest may work, it leave little flexibility between machines and also takes a lot more work then I was hoping to do. actually, it leaves a lot of flexibility between machines. use ed or 'perl -i' scripts to automatically edit config files in place. For example, I have 20 machines at a co location I need to go install. Right now with Red Hat I can take my laptop, slap a floppy in each machine, turn 'em on, 5 minutes later I have 20 fully configured machines ready to rock. you can do the same thing with debian...just install the nfs server package on your laptop. craig -- craig sanders
Re: Mass install / Autoinstall (Was: Re: Debian vs Red Hat??? I need info.)
At 09:55 PM 5/17/00 -0700, Karl M. Hegbloom wrote: copy everything from the master drive to the copy, then run the appropriate Lilo command to make that copy bootable. You can then mount it in another machine and it's ready to go. You have to filter some things out when you copy. See below. You can't do that, I've tried it before. Lilo can't be installed on any secondary disk. Don't ask me why because I don't know. There's a HOWTO about it. +---+ |-=I T ' S P R I N C I P L E T H A T C O U N T S=- | |=- -=ALAN KEYES FOR PRESIDENT=- -=| | Balanced Budgets Personal Freedoms Morality Lower Tax | |=-- http://www.Keyes2000.com. --=| ++
Re: Mass install / Autoinstall (Was: Re: Debian vs Red Hat??? I need info.)
If kickstart is a red hat package, you can install it on debian using alien. Then you can use red hat's kickstart to install debian. :) At 01:55 PM 5/18/00 -0400, Jeremy Hansen wrote: Most of the answers I've been getting on this subject seem like total hacks, which may work but really are tricks to doing this. I was really looking for something within debian that's built to do kickstart type installations. Although what you suggest may work, it leave little flexibility between machines and also takes a lot more work then I was hoping to do. Only for the initial setup. Once your base install is made, a few scripts written, it can become 100% automatic. It's just not 100% automatic out of the box. +---+ |-=I T ' S P R I N C I P L E T H A T C O U N T S=- | |=- -=ALAN KEYES FOR PRESIDENT=- -=| | Balanced Budgets Personal Freedoms Morality Lower Tax | |=-- http://www.Keyes2000.com. --=| ++
Re: Mass install / Autoinstall (Was: Re: Debian vs Red Hat??? I need info.)
Hmm, I don't agree here. Kickstart is a way of automating the tasks already involved with a manual install. It does what it's supposed to do quite well and actually with the flexibility available, I rarely encounter a situation that requires more custom things. Hacks can be included in kickstart during the %post procedure where you can basically write your script to do whatever. I've been using Linux long enough that I don't need to use the hacker way around things for all purposes. For me it's the bottom line. Kickstart lets me setup a lot of machines very quickly with pretty much limitless control over each install. Kickstart is part of anaconda and it is design for what it does, slapping cpio tar and all the other tools you can pass an argument to is just a mess. -jeremy On Thu, May 18, 2000 at 01:55:37PM -0400, Jeremy Hansen wrote: Most of the answers I've been getting on this subject seem like total hacks, which may work but really are tricks to doing this. I was really looking for something within debian that's built to do kickstart type installations. huh? what do you think kickstart is? it's the same kind of total hack - the difference is that you have to do it RedHat's way whether you like it or not, and it pretends to be easy enough to use that you don't need to know what you're doing to run it. personally, i think that anyone who needs to mass-build machines *SHOULD* know exactly what they are doing. i wouldn't trust any machine built by someone who needed such point-and-click tools. Although what you suggest may work, it leave little flexibility between machines and also takes a lot more work then I was hoping to do. actually, it leaves a lot of flexibility between machines. use ed or 'perl -i' scripts to automatically edit config files in place. For example, I have 20 machines at a co location I need to go install. Right now with Red Hat I can take my laptop, slap a floppy in each machine, turn 'em on, 5 minutes later I have 20 fully configured machines ready to rock. you can do the same thing with debian...just install the nfs server package on your laptop. craig -- craig sanders -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED] -- http://www.xxedgexx.com | [EMAIL PROTECTED] -
Re: Mass install / Autoinstall (Was: Re: Debian vs Red Hat??? I need info.)
Well it's funny you brought that up because I was considering just making one huge rpm of debian and then using kickstart. Kickstart is a part of Red Hat's install, Anaconda, not really an rpm but I get your point. -jeremy If kickstart is a red hat package, you can install it on debian using alien. Then you can use red hat's kickstart to install debian. :) At 01:55 PM 5/18/00 -0400, Jeremy Hansen wrote: Most of the answers I've been getting on this subject seem like total hacks, which may work but really are tricks to doing this. I was really looking for something within debian that's built to do kickstart type installations. Although what you suggest may work, it leave little flexibility between machines and also takes a lot more work then I was hoping to do. Only for the initial setup. Once your base install is made, a few scripts written, it can become 100% automatic. It's just not 100% automatic out of the box. +---+ |-=I T ' S P R I N C I P L E T H A T C O U N T S=- | |=- -=ALAN KEYES FOR PRESIDENT=- -=| | Balanced Budgets Personal Freedoms Morality Lower Tax | |=-- http://www.Keyes2000.com. --=| ++ -- http://www.xxedgexx.com | [EMAIL PROTECTED] -
Re: Mass install / Autoinstall (Was: Re: Debian vs Red Hat??? I need info.)
Craig == Craig Sanders [EMAIL PROTECTED] writes: For example, I have 20 machines at a co location I need to go install. Right now with Red Hat I can take my laptop, slap a floppy in each machine, turn 'em on, 5 minutes later I have 20 fully configured machines ready to rock. Craig you can do the same thing with debian...just install the nfs server Craig package on your laptop. I think that with `Woody' we'll have something as good as or better than KickStart. Read up on `debconf', and think about what I said about creating a custom Debian `baseX_X.tgz'. -- Those who do not study Lisp are doomed to reimplement it - Poorly. A few months in the laboratory often saves several hours at the library. mailto:[EMAIL PROTECTED] (Karl M. Hegbloom)