Hello,
I just got back from a datacenter where I installed a new server running CentOS 5.3. I set up a software raid with md0 being /. (Only one mount, didn't split anything up.)
I just discovered that instead of raid 1, it did raid 0... Not exactly what I wanted! Probably missed that little check when installing. So now, I need to change this. But since the whole OS is on this raided partition that I want to change, I'm a bit stuck.
Is there a way to launch an installer while in the OS that I can re-install the whole thing over the internet, without having to physically go to the box? Speed of installation is not that much of an issue. It would obviously need to keep up some sort of shell with my current eth0 settings during the install.
Thanks for any ideas!
Bob
Bob Puff wrote:
Is there a way to launch an installer while in the OS that I can re-install the whole thing over the internet, without having to physically go to the box? Speed of installation is not that much of an issue. It would obviously need to keep up some sort of shell with my current eth0 settings during the install.
you'd need a remote console of some sort on that box. many brand name servers have these, HP calls it iLO, Dell calls it DRAC, etc. these have their own ethernet port, which has to be connected to the network and configured. you'd also need a PXE server on the same network segment to do a jumpstart install from, or use a virtual CD with that remote console (slow if you're not onsite)
Barring that, I think you are SOL.
On Sat, Jun 13, 2009 at 10:18 AM, John R Piercepierce@hogranch.com wrote:
Bob Puff wrote:
Is there a way to launch an installer while in the OS that I can re-install the whole thing over the internet, without having to physically go to the box? Speed of installation is not that much of an issue. It would obviously need to keep up some sort of shell with my current eth0 settings during the install.
you'd need a remote console of some sort on that box. many brand name servers have these, HP calls it iLO, Dell calls it DRAC, etc. these have their own ethernet port, which has to be connected to the network and configured. you'd also need a PXE server on the same network segment to do a jumpstart install from, or use a virtual CD with that remote console (slow if you're not onsite)
Barring that, I think you are SOL.
John: If the DC has KVM over IP, they can connect to his box, can he do it with that? Lanny
Lanny Marcus wrote:
John: If the DC has KVM over IP, they can connect to his box, can he do it with that? Lanny
sure, but I've not seen many ISP colocs who offer that... and that would still leave the issue of setting up a PXE server on the same network segment (at least, the DHCP part of PXE has to be on the same segment, the TFTP and NFS servers can in theory be elsewhere)
I just got back from a datacenter where I installed a new server running CentOS 5.3. I set up a software raid with md0 being /. (Only one mount, didn't split anything up.)
How about /boot? Where's that?
Although I have never done it, you can boot anaconda from grub and I wonder if you couldn't put a ks= as parameter in the grub conf? Setting up software raid in ks is easy. You can also install via url with ks.
Really, you have nothing to lose :)
jlc
Although I have never done it, you can boot anaconda from grub and I wonder if you couldn't put a ks= as parameter in the grub conf? Setting up software raid in ks is easy. You can also install via url with ks.
I just did this and it works. A complete grub stanza booted and installed via url with a ks file completely non interactively without a dhcp/tftp server or cd.
"Works For Me" (tm) Heh, jlc
On Sat, Jun 13, 2009 at 4:00 PM, Joseph L. CasaleJCasale@activenetwerx.com wrote:
Although I have never done it, you can boot anaconda from grub and I wonder if you couldn't put a ks= as parameter in the grub conf? Setting up software raid in ks is easy. You can also install via url with ks.
I just did this and it works. A complete grub stanza booted and installed via url with a ks file completely non interactively without a dhcp/tftp server or cd.
"Works For Me" (tm)
Karan has documented this for centos4, though it's easy enough to do the same/similar thing for centos5. Have a read over http://www.karan.org/blog/index.php/2005/06/15/upgrading_to_centos4_over_a_r... and that should show you what you want.
On 06/13/2009 07:56 AM, Bob Puff wrote:
Is there a way to launch an installer while in the OS that I can re-install the whole thing over the internet, without having to physically go to the box?
No, but you can:
1) Grab the vmlinuz and initrd.img from the installation CD's images/pxeboot directory and save them to /boot/ 2) Set up a stanza in grub.conf that will load that vmlinuz and initrd. You will need to list the "ksdevice", ip address, netmask, gateway, and kickstart URL in grub.conf. 3) Create a kickstart file that specifies exactly the configuration that you require and store it at the URL you used in grub.conf.
Test the entire process on a local machine until it works exactly the way that you want it. You can use "system-config-kickstart" to set up your kickstart file. It would be best to use that tool on CentOS 5. Newer systems (like Fedora 11) use an updated syntax that may not parse correctly on CentOS 5.
I just got back from a datacenter where I installed a new server running CentOS 5.3. I set up a software raid with md0 being /. (Only one mount, didn't split anything up.)
I just discovered that instead of raid 1, it did raid 0... Not exactly what I wanted! Probably missed that little check when installing. So now, I need to change this. But since the whole OS is on this raided partition that I want to change, I'm a bit stuck.
Is there a way to launch an installer while in the OS that I can re-install the whole thing over the internet, without having to physically go to the box? Speed of installation is not that much of an issue. It would obviously need to keep up some sort of shell with my current eth0 settings during the install.
install a small boot kernel that drops straight to anaconda and in that have ip=xxxx gateway=xxxx ks=http://xxxx/foo.ks
have you install tree available and providing there are no mistakes in your ks, test it first!, then a reinstall should happen this time with the software raid
on 6-13-2009 7:56 AM Bob Puff spake the following:
Hello,
I just got back from a datacenter where I installed a new server running CentOS 5.3. I set up a software raid with md0 being /. (Only one mount, didn't split anything up.)
I just discovered that instead of raid 1, it did raid 0... Not exactly what I wanted! Probably missed that little check when installing. So now, I need to change this. But since the whole OS is on this raided partition that I want to change, I'm a bit stuck.
Is there a way to launch an installer while in the OS that I can re-install the whole thing over the internet, without having to physically go to the box? Speed of installation is not that much of an issue. It would obviously need to keep up some sort of shell with my current eth0 settings during the install.
Thanks for any ideas!
Bob
If you can manage to free a small amount of space in a non raid 0 boot partition you can do a remote install. You can also do one with the netboot cd if your colo is willing to place a cd and reboot for you. You can do a vnc remote install and pull everything from the internet if the box has access.
If you don't have any space that WON'T need to be formatted during the install, you might just be out of luck unless you can get a boot cd to them.