CobblerTest
From WBITT's Cooker!
(→=) |
(→Cobbler install on Dell 1950 Servers) |
||
| Line 26: | Line 26: | ||
* reposync (yum-utils provides reposync) | * reposync (yum-utils provides reposync) | ||
| + | <pre> | ||
yum -y install httpd yum-utils tftp-server xinetd dhcp | yum -y install httpd yum-utils tftp-server xinetd dhcp | ||
| + | </pre> | ||
Make sure you are not running kernel-xen on your base OS. (Although it does not matter, but I just want to keep things simple). My scenario requires me not to run kernel-xen. (We are not using virtualization features. These are pure compute nodes.) | Make sure you are not running kernel-xen on your base OS. (Although it does not matter, but I just want to keep things simple). My scenario requires me not to run kernel-xen. (We are not using virtualization features. These are pure compute nodes.) | ||
Revision as of 22:34, 10 February 2010
Contents |
Cobbler Howto
Cobbler install on Dell 1950 Servers
This setup consists of four Dell PowerEdge 1950 servers, connected to each other in a cluster form. The intended purpose of this excercise is to find a way to provision compute nodes, without hassle. This should be ideal for data centers (both HPCC and non-HPCC).
Help from:-
http://www.openskill.info/blog.php?ID=16 http://magazine.redhat.com/2007/08/10/cobbler-how-to-set-up-a-network-boot-server-in-10-minutes/ http://www.enterprisenetworkingplanet.com/netsysm/article.php/3773596 http://www.kanarip.com/courses/Deploying_Linux_Using_Cobbler/
The system has Centos 5.4 x86_64 installed as base OS. You might want to update it before continuing.
yum -y update
After this you should make sure that you have the following installed, from your distribution media / repository.
- httpd
- tftp-server
- xinetd
- reposync (yum-utils provides reposync)
yum -y install httpd yum-utils tftp-server xinetd dhcp
Make sure you are not running kernel-xen on your base OS. (Although it does not matter, but I just want to keep things simple). My scenario requires me not to run kernel-xen. (We are not using virtualization features. These are pure compute nodes.)
Using the link: https://fedorahosted.org/cobbler/wiki/DownloadInstructions
Setup Cobbler Repositories:- (Needs internet connection).
[root@headnode ~]# vi /etc/yum.repos.d/CentOS-Base.repo . . . . . . [cobbler-stable] name=Cobbler-Stable baseurl=http://download.fedora.redhat.com/pub/epel/5/x86_64 [cobbler-testing] name=Cobbler-Testing baseurl=http://download.fedora.redhat.com/pub/epel/testing/5/x86_64 . . . . . . [root@headnode ~]#
Now, install cobbler:-
[root@headnode ~]# yum -y install cobbler koan cman
I have various CD images in my /data/cdimages location. I expanded the ISO image and copied it in CentOS-5.4-x86_64 .
[root@headnode cdimages]# ls CentOS-5.3-i386.DVD.iso CentOS-5.4-x86_64 CentOS-5.4-x86_64-bin-DVD.iso [root@headnode cdimages]#
Assuming you have entire CD media copied in your /data/cdimages/CentOS-5.4-x86_64 , you are now ready to import the repository.
cobbler import --mirror=/data/cdimages/CentOS-5.4-x86_64 --name=Centos5.4-x86_64
or
cobbler import --path=/data/cdimages/CentOS-5.4-x86_64 --name=Centos5.4-x86_64
During import you should get some screen out put as:
[root@headnode cdimages]# cobbler import --mirror=/data/cdimages/CentOS-5.4-x86_64 --name=Centos5.4-x86_64
. . .
. . .
repodata/primary.sqlite.bz2
2141520 100% 3.07MB/s 0:00:00 (xfer#3468, to-check=2/3478)
repodata/primary.xml.gz
1189512 100% 1.67MB/s 0:00:00 (xfer#3469, to-check=1/3478)
repodata/repomd.xml
2142 100% 3.09kB/s 0:00:00 (xfer#3470, to-check=0/3478)
sent 4333936317 bytes received 76408 bytes 39580024.89 bytes/sec
total size is 4333090152 speedup is 1.00
---------------- (adding distros)
- found content (breed=redhat) at /var/www/cobbler/ks_mirror/Centos5.4-x86_64/images/xen
- creating new distro: Centos5.4-xen-x86_64
- creating new profile: Centos5.4-xen-x86_64
- found content (breed=redhat) at /var/www/cobbler/ks_mirror/Centos5.4-x86_64/images/pxeboot
- creating new distro: Centos5.4-x86_64
- creating new profile: Centos5.4-x86_64
- creating new profile: rescue-Centos5.4-x86_64
---------------- (associating repos)
- traversing distro Centos5.4-xen-x86_64
- found content (breed=redhat) at /var/www/cobbler/ks_mirror/Centos5.4-x86_64/images/xen
- descent into /var/www/cobbler/ks_mirror/Centos5.4-x86_64
- processing repo at : /var/www/cobbler/ks_mirror/Centos5.4-x86_64
- need to process repo/comps: /var/www/cobbler/ks_mirror/Centos5.4-x86_64
- looking for /var/www/cobbler/ks_mirror/Centos5.4-x86_64/repodata/*comps*.xml
- createrepo -c cache --groupfile /var/www/cobbler/ks_mirror/Centos5.4-x86_64/repodata/comps.xml /var/www/cobbler/ks_mirror/Centos5.4-x86_64
3309/3309 - CentOS/sblim-cim-client-javadoc-1-35.el5.x86_64.rpm rpm64.rpm
Saving Primary metadata
Saving file lists metadata
Saving other metadata
- traversing distro Centos5.4-x86_64
- found content (breed=redhat) at /var/www/cobbler/ks_mirror/Centos5.4-x86_64/images/pxeboot
- descent into /var/www/cobbler/ks_mirror/Centos5.4-x86_64
- processing repo at : /var/www/cobbler/ks_mirror/Centos5.4-x86_64
- need to process repo/comps: /var/www/cobbler/ks_mirror/Centos5.4-x86_64
- looking for /var/www/cobbler/ks_mirror/Centos5.4-x86_64/repodata/*comps*.xml
- createrepo -c cache --groupfile /var/www/cobbler/ks_mirror/Centos5.4-x86_64/repodata/comps.xml /var/www/cobbler/ks_mirror/Centos5.4-x86_64
3309/3309 - CentOS/sblim-cim-client-javadoc-1-35.el5.x86_64.rpm rpm64.rpm
Saving Primary metadata
Saving file lists metadata
Saving other metadata
---------------- (associating kickstarts)
- found content (breed=redhat) at /var/www/cobbler/ks_mirror/Centos5.4-x86_64/images/pxeboot
- found content (breed=redhat) at /var/www/cobbler/ks_mirror/Centos5.4-x86_64/images/xen
- found content (breed=redhat) at /var/www/cobbler/ks_mirror/Centos5.4-x86_64/images/pxeboot
[root@headnode cdimages]#
Time for cobbler configurations:-
[root@headnode ~]# vi /etc/cobbler/settings server = 192.168.1.203 nextserver=192.168.1.203 pxe_just_once = 1 . . .
Now you should start the cobblerd service along httpd, dhcpd and tftp server.
chkconfig --level 35 cobblerd on chkconfig --level 35 httpd on chkconfig --level 35 xinetd on chkconfig --level 35 dhcpd on service cobblerd restart service httpd restart service dhcpd restart service xinetd restart
Should enable the tftp service in /etc/xinetd.d/tftp , and change disabled=no .
Now run cobbler check and fix any obvious errors.
cobbler check # important
You may want to disable un-necessary services on this server before moving on:
chkconfig --level 35 NetworkManager off chkconfig --level 35 autofs off chkconfig --level 35 avahi-daemon off chkconfig --level 35 avahi-dnsconfd off chkconfig --level 35 bluetooth off chkconfig --level 35 cups off chkconfig --level 35 dc_client off chkconfig --level 35 dc_server off chkconfig --level 35 dhcrelay off chkconfig --level 35 dnsmasq off chkconfig --level 35 dund off chkconfig --level 35 firstboot off chkconfig --level 35 gpm off chkconfig --level 35 hidd off chkconfig --level 35 ip6tables off chkconfig --level 35 iptables off chkconfig --level 35 irda off chkconfig --level 35 iscsi off chkconfig --level 35 iscsid off chkconfig --level 35 kdump off chkconfig --level 35 libvirtd off chkconfig --level 35 lisa off chkconfig --level 35 lvm2-monitor off chkconfig --level 35 mdmonitor off chkconfig --level 35 mdmpd off chkconfig --level 35 multipathd off chkconfig --level 35 mysqld off chkconfig --level 35 named off chkconfig --level 35 netconsole off chkconfig --level 35 netfs off chkconfig --level 35 netplugd off chkconfig --level 35 nfs off chkconfig --level 35 nfslock off chkconfig --level 35 nscd off chkconfig --level 35 ntpd off chkconfig --level 35 oddjobd off chkconfig --level 35 pand off chkconfig --level 35 pcscd off chkconfig --level 35 psacct off chkconfig --level 35 qemu off chkconfig --level 35 rawdevices off chkconfig --level 35 rdisc off chkconfig --level 35 readahead_later off chkconfig --level 35 restorecond off chkconfig --level 35 rpcgssd off chkconfig --level 35 rpcidmapd off chkconfig --level 35 rpcsvcgssd off chkconfig --level 35 saslauthd off chkconfig --level 35 sendmail off chkconfig --level 35 setroubleshoot off chkconfig --level 35 smartd off chkconfig --level 35 smb off chkconfig --level 35 squid off chkconfig --level 35 tcsd off chkconfig --level 35 tux off chkconfig --level 35 vncserver off chkconfig --level 35 wdaemon off chkconfig --level 35 winbind off chkconfig --level 35 wpa_supplicant off chkconfig --level 35 xend off chkconfig --level 35 xendomains off chkconfig --level 35 ypbind off chkconfig --level 35 yum-updatesd off
End of day 1.
Day 2
[root@headnode ~]# cobbler check The following potential problems were detected: #0: service dhcpd is not running #1: since iptables may be running, ensure 69, 80, and 25151 are unblocked #2: yum-utils need to be at least version 1.1.17 for reposync -l, current version is 1.1.16 #3: The default password used by the sample templates for newly installed machines (default_password_crypted in /etc/cobbler/settings) is still set to 'cobbler' and should be changed [root@headnode ~]#
As you can see, 1-3 are more important to check. 4 is less important.
Check various lists from cobbler. These are created by default when you import the repository for the first time.
[root@headnode ~]# cobbler distro list Centos5.4-x86_64 Centos5.4-xen-x86_64 [root@headnode ~]# [root@headnode ~]# cobbler profile list Centos5.4-x86_64 Centos5.4-xen-x86_64 rescue-Centos5.4-x86_64 [root@headnode ~]#
Currently. No systems are registered with cobbler at the moment. As you can see below:
[root@headnode ~]# cobbler system list [root@headnode ~]#
First, create a (separate) profile for compute-nodes:-
[root@headnode ~]# cobbler profile copy --name=Centos5.4-x86_64 --newname=ComputeNodes
[root@headnode ~]# cobbler profile list
Centos5.4-x86_64
Centos5.4-xen-x86_64
ComputeNodes # <------------ Newly added profile!
rescue-Centos5.4-x86_64
[root@headnode ~]#
[root@headnode ~]# cobbler profile report --name=ComputeNodes
profile : ComputeNodes
distro : Centos5.4-x86_64
comment :
created : Tue Feb 2 09:06:50 2010
dhcp tag : default
enable menu : True
kernel options : {}
kickstart : /var/lib/cobbler/kickstarts/default.ks # < ------------ Notice this default Kickstart.
ks metadata : {}
mgmt classes : []
modified : Tue Feb 2 09:06:50 2010
name servers : []
name servers search : []
owners : ['admin']
post kernel options : {}
redhat mgmt key : <<inherit>>
redhat mgmt server : <<inherit>>
repos : []
server : <<inherit>>
template_files : {}
virt bridge : xenbr0
virt cpus : 1
virt file size : 5
virt path :
virt ram : 512
virt type : qemu
[root@headnode ~]#
The name of default.ks needs to be changed, or fill default.ks with your kickstart, because it is empty by default.
[root@headnode ~]# cat /var/lib/cobbler/kickstarts/default.ks # this file intentionally left blank # admins: edit it as you like, or leave it blank for non-interactive install
There is a sample.ks, which we will use for our compute nodes.
[root@headnode ~]# cat /var/lib/cobbler/kickstarts/sample.ks
#platform=x86, AMD64, or Intel EM64T
# System authorization information
auth --useshadow --enablemd5
# System bootloader configuration
bootloader --location=mbr
# Partition clearing information
clearpart --all --initlabel
# Use text mode install
text
# Firewall configuration
firewall --enabled
# Run the Setup Agent on first boot
firstboot --disable
# System keyboard
keyboard us
# System language
lang en_US
# Use network installation
url --url=$tree
# If any cobbler repo definitions were referenced in the kickstart profile, include them here.
$yum_repo_stanza
# Network information
$SNIPPET('network_config')
# Reboot after installation
reboot
#Root password
rootpw --iscrypted $default_password_crypted
# SELinux configuration
selinux --disabled
# Do not configure the X Window System
skipx
# System timezone
timezone America/New_York
# Install OS instead of upgrade
install
# Clear the Master Boot Record
zerombr
# Allow anaconda to partition the system as needed
autopart
%pre
$SNIPPET('log_ks_pre')
$kickstart_start
$SNIPPET('pre_install_network_config')
# Enable installation monitoring
$SNIPPET('pre_anamon')
%packages
$SNIPPET('func_install_if_enabled')
%post
$SNIPPET('log_ks_post')
# Start yum configuration
$yum_config_stanza
# End yum configuration
$SNIPPET('post_install_kernel_options')
$SNIPPET('post_install_network_config')
$SNIPPET('func_register_if_enabled')
$SNIPPET('download_config_files')
$SNIPPET('koan_environment')
$SNIPPET('redhat_register')
$SNIPPET('cobbler_register')
# Enable post-install boot notification
$SNIPPET('post_anamon')
# Start final steps
$kickstart_done
# End final steps
[root@headnode ~]#
Copy the sample file as Compute.ks :
[root@headnode ~]# cd /var/lib/cobbler/kickstarts/ [root@headnode kickstarts]# cp sample.ks Compute.ks
Edit the kickstart file Compute.ks and make necessary changes:
[root@headnode kickstarts]# vi Compute.ks
#platform=x86, AMD64, or Intel EM64T
# System authorization information
auth --useshadow --enablemd5
# System bootloader configuration
bootloader --location=mbr
# Partition clearing information
clearpart --all --initlabel
# Use text mode install
text
# Firewall configuration
firewall --disabled
# Run the Setup Agent on first boot
firstboot --disable
# System keyboard
keyboard us
# System language
lang en_US
# Use network installation
url --url=$tree
# If any cobbler repo definitions were referenced in the kickstart profile, include them here.
$yum_repo_stanza
# Network information
$SNIPPET('network_config')
# Reboot after installation
reboot
#Root password
rootpw --iscrypted $default_password_crypted
# SELinux configuration
selinux --disabled
# Do not configure the X Window System
skipx
# System timezone
timezone Asia/Riyadh
# Install OS instead of upgrade
install
# Clear the Master Boot Record
zerombr
# Allow anaconda to partition the system as needed
# autopart
part /boot --size 512
part swap --size 1024
part / --size 4096
%pre
$SNIPPET('log_ks_pre')
$kickstart_start
$SNIPPET('pre_install_network_config')
# Enable installation monitoring
$SNIPPET('pre_anamon')
%packages
@core
$SNIPPET('func_install_if_enabled')
%post
$SNIPPET('log_ks_post')
# Start yum configuration
$yum_config_stanza
# End yum configuration
$SNIPPET('post_install_kernel_options')
$SNIPPET('post_install_network_config')
$SNIPPET('func_register_if_enabled')
$SNIPPET('download_config_files')
$SNIPPET('koan_environment')
$SNIPPET('redhat_register')
$SNIPPET('cobbler_register')
# Enable post-install boot notification
$SNIPPET('post_anamon')
# Start final steps
$kickstart_done
# End final steps
[root@headnode kickstarts]#
Now, assign the new kickstart to the profile we just created/copied:
[root@headnode kickstarts]# cobbler profile edit --name=ComputeNodes --kickstart=/var/lib/cobbler/kickstarts/Compute.ks
Verify the change:-
[root@headnode kickstarts]# cobbler profile report --name=ComputeNodes | grep kickstart kickstart : /var/lib/cobbler/kickstarts/Compute.ks [root@headnode kickstarts]#
Define Systems
Now define/ add compute nodes. Remember, that my compute nodes have an IPMI capable interface, which have an out-of-band IP (separate network IP), than the access / production network. Right now I am only adding systems with the most basic information:
[root@headnode ~]# cobbler system add --name=compute1 --profile=ComputeNodes --mac=00:1D:09:6A:27:91 [root@headnode ~]# cobbler system add --name=compute2 --profile=ComputeNodes --mac=00:1D:09:6B:06:81
Note: In case you try to make a duplicate entry:-
[root@headnode ~]# cobbler system add --name=compute1 --profile=ComputeNodes --mac=00:1D:09:6B:06:7F An object already exists with that name. Try 'edit'?
Check cobbler system list:
[root@headnode ~]# cobbler system list compute1 compute2 [root@headnode ~]#
Looks good. You can check report of the nodes now. I am only interested in few parameters.
[root@headnode ~]# cobbler system report | egrep -i "system|profile|power type|interface|mac" system : compute1 profile : ComputeNodes power type : ipmitool interface : eth0
mac address : 00:1D:09:6A:27:8F
system : compute2 profile : ComputeNodes power type : ipmitool interface : eth0
mac address : 00:1D:09:6B:06:7F
[root@headnode ~]#
I need to change the interface from eth0 to eth1, because my nodes are connected to access network through eth1. Whereas eth1 is the management / IPMI / BMC network.
I can simply delete these two systems and recreate them without a hassle:
[root@headnode ~]# cobbler system add --name=compute1 --profile=ComputeNodes --mac=00:1D:09:6A:27:91 --hostname=compute1.example.com --interface=eth1 --static=1 --ip=192.168.1.1 --subnet=255.255.255.0 --kopts=noipv6 --power-address=192.168.0.1 --power-user=root --power-pass=ipmisecret [root@headnode ~]# cobbler system add --name=compute2 --profile=ComputeNodes --mac=00:1D:09:6B:06:81 --hostname=compute2.example.com --interface=eth1 --static=1 --ip=192.168.1.2 --subnet=255.255.255.0 --kopts=noipv6 --power-address=192.168.0.2 --power-user=root --power-pass=ipmisecret [root@headnode ~]# Note: The BMC interface has the user name "root" and password "ipmisecret", are defined in each node's BMC configuration in BIOS.
Now once you modify anything in the cobbler system, you will need to do a "cobbler sync", which will make necessary changes in kickstart and tftp files and restart dhcpd. (I wasted one full day browsing the net, to figure out this one!)
[root@headnode ~]# cobbler sync Shutting down dhcpd: [ OK ] Starting dhcpd: [ OK ] [root@headnode ~]#
And if you make any changes in cobbler's /etc/cobbler/settings file or any of it's subsystem, you will need to restart cobblerd service.
service cobblerd restart cobbler sync
If you look at the TFTP file created by cobbler for compute1, it looks like this:
[root@headnode ~]# cat /tftpboot/pxelinux.cfg/01-00-1d-09-6a-27-8f
default linux
prompt 0
timeout 1
label linux
kernel /images/Centos5.4-x86_64/vmlinuz
ipappend 2
append initrd=/images/Centos5.4-x86_64/initrd.img ksdevice=bootif lang= kssendmac text ks=http://192.168.1.200/cblr/svc/op/ks/system/compute1
[root@headnode ~]#
<pre>
The resultant dhcpd.conf file looks like this:
<pre>
cat /etc/dhcpd.conf
# ******************************************************************
# Cobbler managed dhcpd.conf file
# generated from cobbler dhcp.conf template (Tue Feb 2 11:47:35 2010)
# Do NOT make changes to /etc/dhcpd.conf. Instead, make your changes
# in /etc/cobbler/dhcp.template, as /etc/dhcpd.conf will be
# overwritten.
# ******************************************************************
ddns-update-style interim;
allow booting;
allow bootp;
ignore client-updates;
set vendorclass = option vendor-class-identifier;
subnet 192.168.1.0 netmask 255.255.255.0 {
option routers 192.168.1.200;
option domain-name-servers 192.168.1.200;
option subnet-mask 255.255.255.0;
range dynamic-bootp 192.168.1.1 192.168.1.10;
filename "/pxelinux.0";
default-lease-time 21600;
max-lease-time 43200;
next-server 192.168.1.200;
}
# group for Cobbler DHCP tag: default
group {
host generic1 {
hardware ethernet 00:1D:09:6B:06:91;
filename "/pxelinux.0";
next-server 192.168.1.200;
}
}
[root@headnode ~]#
For two nodes, I had the following TFTP files:
[root@headnode ~]# ls /tftpboot/pxelinux.cfg/ -l total 12 -rw-r--r-- 1 root root 272 Feb 8 13:18 01-00-1d-09-6a-27-91 -rw-r--r-- 1 root root 272 Feb 8 13:18 01-00-1d-09-6b-06-81 -rw-r--r-- 1 root root 1091 Feb 8 13:18 default [root@headnode ~]#
If your systems have IPMI capable interfaces, then you must make sure that you have ipmitools, ipmiconsole and ipmipower installed.
[root@headnode ~]# yum -y install OpenIPMI-tools freeipmi-* [root@headnode ~]# rpm -ql OpenIPMI-tools /etc/rc.d/init.d/ipmievd /etc/sysconfig/ipmievd /usr/bin/ipmitool
Time to reboot these nodes and see if they are able to install. However, make sure that their netboot switch is enabled.
[root@headnode power]# cobbler system edit --name=compute1 --netboot=1 [root@headnode power]# cobbler system edit --name=compute2 --netboot=1 [root@headnode power]# cobbler report --what=systems | egrep -i "hostname|netboot" hostname : compute1.example.com netboot enabled? : True hostname : compute2.example.com netboot enabled? : True [root@headnode power]#
There is no "install" command. Instead all the installation mechanism has been made ready. You need to reboot the nodes to get installed.
[root@headnode power]# cobbler system reboot --name=compute1
cobbler power configuration is:
type : ipmitool
address: 192.168.0.1
user : root
id :
- /usr/bin/ipmitool -H 192.168.0.1 -U root -P ipmisecret power off
Chassis Power Control: Down/Off
cobbler power configuration is:
type : ipmitool
address: 192.168.0.1
user : root
id :
- /usr/bin/ipmitool -H 192.168.0.1 -U root -P ipmisecret power on
Chassis Power Control: Up/On
[root@headnode power]#
If, after reboot of node, the installation starts and finishes and starts again, it means your nodes are stuck in an install loop. This happens, because PXE is the first boot device in the nodes, which is the required boot sequence in HPCC scenarios. What you need to do is to make sure that a particular setting is enabled in the /etc/cobblet/settings file. Without setting that, the netboot flag will remain switched on in cobbler configuration and the nodes will keep rebooting and keep installing.
[root@headnode power]# cobbler report --what=systems | egrep -i "hostname|netboot" hostname : compute1.example.com netboot enabled? : True hostname : compute2.example.com netboot enabled? : True [root@headnode power]#
For such situations, there is a flag in /etc/cobbler/settings, which turns off the netboot flag, when the installation is done. It is "pxe_just_once" and is set to 0 by default. It needs to be turned to 1 in this situation.
pxe_just_once: 1
Day 3. Few other issues.
My nodes installed properly and booted properly. However they were taking too much time in trying to obtain IP for eth0 interface through dhcp. Since there is no dhcp on eth0 and I don't need to enable the eth0 interface in OS, because it is enabled and properly configured in the BIOS already, I had to find a way to fix that. Here are four items which I wanted to know the solution:
1) Static IP for eth1. Means, after installation of a node is complete, it should have a static IP configured in it instead of asking that from DHCP server. 2) Disable eth0, so it does not wait for dhcp when the node reboots from local OS/disk. 3) Console redirection during TFTP and installation. I want to see the console of the node on my headnode, instead of manually seeing it on the KVM. (It does not make sense to run to the data center everytime you need to look at screen of a node.) 4) How to check if the nodes are installed? Pretty important!
1) Static IP:-
Simply add a --static=1 to either the cobbler system add command, or modify an already defined system with this setting.
[root@headnode ~]# cobbler system edit --name compute1 --interface=eth1 --static=1
This is the solution for setting static IP, post-reboot. The installation got done through PXE (good).
Here is what cobbler did, after the node is installed and rebooted from local disk:-
[root@compute1 ~]# cat /etc/sysconfig/network-scripts/ifcfg-eth1 DEVICE=eth1 HWADDR=00:1D:09:6A:27:91 ONBOOT=yes BOOTPROTO=static IPADDR=192.168.1.1 NETMASK=255.255.255.0 [root@compute1 ~]#
This is exactly what I wanted. Good.
2) Disable eth0. For this, I used the same technique. I just mentioned that it is also static, and did not provide any IP for it. That is because the (out of band) IP is already configured in the node's BIOS.
[root@headnode ~]# cobbler system edit --name compute1 --interface=eth0 --static=1 [root@headnode ~]# cobbler report system --name=compute1 system : compute1 profile : ComputeNodes comment : . . . interface : eth0 mac address : is static? : True ip address : subnet : static routes : [] dhcp tag : interface : eth1 mac address : 00:1D:09:6A:27:91 is static? : True ip address : 192.168.1.1 subnet : 255.255.255.0 static routes : [] dhcp tag : [root@headnode ~]#
Note that --delete-interface did NOT do the trick.
[root@headnode ~]# cobbler system edit --name=compute1 --delete-interface=eth0 This did not work.
This time it worked. I have lovely results after the nodes are installed and rebooted!
[root@compute1 ~]# ifconfig eth0 Link encap:Ethernet HWaddr 00:1D:09:6A:27:8F
inet6 addr: fe80::21d:9ff:fe6a:278f/64 Scope:Link
eth1 Link encap:Ethernet HWaddr 00:1D:09:6A:27:91
inet addr:192.168.1.1 Bcast:192.168.1.255 Mask:255.255.255.0
inet6 addr: fe80::21d:9ff:fe6a:2791/64 Scope:Link
lo Link encap:Local Loopback
inet addr:127.0.0.1 Mask:255.0.0.0
inet6 addr: ::1/128 Scope:Host
[root@compute1 ~]# cat /etc/sysconfig/network-scripts/ifcfg-eth0 DEVICE=eth0 HWADDR=00:1D:09:6A:27:8F ONBOOT=yes BOOTPROTO=none
[root@compute1 ~]# cat /etc/sysconfig/network-scripts/ifcfg-eth1 DEVICE=eth1 HWADDR=00:1D:09:6A:27:91 ONBOOT=yes BOOTPROTO=static IPADDR=192.168.1.1 NETMASK=255.255.255.0 [root@compute1 ~]#
Notice that Cobbler set BOOTPROTO=none for eth0.
</pre>
However the node WILL NOT ping any other node on IPMI interface. This is only controllable/ do-able in one direction, from headnode towards compute node. Remember that the headnode's management interface HAS an out of band IP configured in the OS.
Thus, don't be alarmed if you see the following, while you are logged in to any of the compute nodes:
[root@compute1 ~]# ping 192.168.0.200 connect: Network is unreachable [root@compute1 ~]# ping 192.168.0.201 connect: Network is unreachable [root@compute1 ~]#
End of day 3.
Day 4. Console redirection.
For (3) We set certain parameters in the distro and profile options.
[root@headnode ~]# cobbler distro list Centos5.4-x86_64 Centos5.4-xen-x86_64 [root@headnode ~]#
Centos5.4-x86_64 is the distro we are using for provisioning the nodes.
[root@headnode ~]# cobbler distro report --name=Centos5.4-x86_64
distro : Centos5.4-x86_64
architecture : x86_64
breed : redhat
created : Mon Feb 1 15:06:46 2010
comment : rhel5.4
initrd : /var/www/cobbler/ks_mirror/Centos5.4-x86_64/images/pxeboot/initrd.img
kernel : /var/www/cobbler/ks_mirror/Centos5.4-x86_64/images/pxeboot/vmlinuz
kernel options : {}
ks metadata : {'tree': 'http://@@http_server@@/cblr/links/Centos5.4-x86_64'}
tree build time : Thu Oct 1 19:57:55 2009
modified : Mon Feb 1 15:10:12 2010
mgmt classes : []
os version : rhel5
owners : ['admin']
post kernel options : {}
redhat mgmt key : <<inherit>>
redhat mgmt server : <<inherit>>
template files : {}
[root@headnode ~]#
[root@headnode ~]# cobbler distro edit --name=Centos5.4-x86_64 --kopts="console=ttyS1,115200"
[root@headnode ~]# cobbler distro report --name=Centos5.4-x86_64
distro : Centos5.4-x86_64
architecture : x86_64
breed : redhat
created : Mon Feb 1 15:06:46 2010
comment : rhel5.4
initrd : /var/www/cobbler/ks_mirror/Centos5.4-x86_64/images/pxeboot/initrd.img
kernel : /var/www/cobbler/ks_mirror/Centos5.4-x86_64/images/pxeboot/vmlinuz
kernel options : {'console': 'ttyS1,115200'}
ks metadata : {'tree': 'http://@@http_server@@/cblr/links/Centos5.4-x86_64'}
tree build time : Thu Oct 1 19:57:55 2009
modified : Tue Feb 9 08:24:51 2010
mgmt classes : []
os version : rhel5
owners : ['admin']
post kernel options : {}
redhat mgmt key : <<inherit>>
redhat mgmt server : <<inherit>>
template files : {}
[root@headnode ~]#
Now check the profiles:-
[root@headnode ~]# cobbler profile list Centos5.4-x86_64 Centos5.4-xen-x86_64 ComputeNodes rescue-Centos5.4-x86_64 [root@headnode ~]#
We are interested in profile named "ComputeNodes":-
[root@headnode ~]# cobbler profile report --name=ComputeNodes
profile : ComputeNodes
distro : Centos5.4-x86_64
comment :
created : Tue Feb 2 09:06:50 2010
dhcp tag : default
enable menu : True
kernel options : {}
kickstart : /var/lib/cobbler/kickstarts/Compute.ks
ks metadata : {}
mgmt classes : []
modified : Tue Feb 2 09:16:28 2010
name servers : []
name servers search : []
owners : ['admin']
post kernel options : {}
redhat mgmt key : <<inherit>>
redhat mgmt server : <<inherit>>
repos : []
server : <<inherit>>
template_files : {}
virt bridge : xenbr0
virt cpus : 1
virt file size : 5
virt path :
virt ram : 512
virt type : qemu
[root@headnode ~]#
[root@headnode ~]# cobbler profile edit --name=ComputeNodes --ksmeta="console=ttyS1,115200"
[root@headnode ~]# cobbler sync
Starting dhcpd: [ OK ]
[root@headnode ~]#
Try re-provisioning the node and see if it works.
[root@headnode ~]# cobbler system edit --name compute1 --netboot-enabled=1
[root@headnode ~]# cobbler sync
Shutting down dhcpd: [ OK ]
Starting dhcpd: [ OK ]
[root@headnode ~]# cat /tftpboot/pxelinux.cfg/01-00-1d-09-6a-27-91
default linux
prompt 0
timeout 1
label linux
kernel /images/Centos5.4-x86_64/vmlinuz
ipappend 2
append initrd=/images/Centos5.4-x86_64/initrd.img ksdevice=bootif lang= console=ttyS1,115200 text noipv6 kssendmac ks=http://192.168.1.200/cblr/svc/op/ks/system/compute1
[root@headnode ~]#
Notice the console option in the newly generated TFTP file above.
ipmiconsole
Use the ipmiconsole command to open console of a remote node on the headnode.
[root@headnode ~]# ipmiconsole --hostname=192.168.0.1 --username=root --password=ipmisecret [SOL established]
Note, if you get an error such as below, that means your headnode, does not have BMC interface correctly configured. Assign it an IP from your BMC network and bring the interface up. It should work.
The error:
[root@headnode ~]# ipmiconsole --hostname=192.168.0.1 --username=root --password=ipmisecret ipmiconsole_submit_block: internal system error
The reason:
[root@headnode ~]# ifconfig
eth0 Link encap:Ethernet HWaddr 00:1D:09:69:F3:BB
inet6 addr: fe80::21d:9ff:fe69:f3bb/64 Scope:Link
eth1 Link encap:Ethernet HWaddr 00:1D:09:69:F3:BD
inet addr:192.168.1.200 Bcast:192.168.1.255 Mask:255.255.255.0
lo Link encap:Local Loopback
inet addr:127.0.0.1 Mask:255.0.0.0
inet6 addr: ::1/128 Scope:Host
The fix:
[root@headnode ~]# ifconfig eth0 192.168.0.201 netmask 255.255.255.0 up
[root@headnode ~]# ifconfig
eth0 Link encap:Ethernet HWaddr 00:1D:09:69:F3:BB
inet addr:192.168.0.201 Bcast:192.168.0.255 Mask:255.255.255.0
eth1 Link encap:Ethernet HWaddr 00:1D:09:69:F3:BD
inet addr:192.168.1.200 Bcast:192.168.1.255 Mask:255.255.255.0
inet6 addr: fe80::21d:9ff:fe69:f3bd/64 Scope:Link
lo Link encap:Local Loopback
inet addr:127.0.0.1 Mask:255.0.0.0
inet6 addr: ::1/128 Scope:Host
[root@headnode ~]# ipmiconsole --hostname=192.168.0.1 --username=root --password=ipmisecret
[SOL established]
power on/off
Use ipmipower command to perform power operations on the nodes.
[root@headnode ~]# ipmipower --hostname=192.168.0.1 --username=root --password=ipmisecret --on 192.168.0.1: ok
Also if the node is off, cycle may not work:-
[root@headnode ~]# ipmipower --hostname=192.168.0.1 --username=root --password=ipmisecret --cycle 192.168.0.1: operation invalid [root@headnode ~]# ipmipower --hostname=192.168.0.1 --username=root --password=ipmisecret --off 192.168.0.1: ok [root@headnode ~]# ipmipower --hostname=192.168.0.1 --username=root --password=ipmisecret --on 192.168.0.1: ok [root@headnode ~]#
Problem with console entries
When using console=ttyS1,57600n8r , we get a node install time slowed down 5 times. Without it, the node install time was 1 minute. With it, the node install time reached 5+ minutes.
Using 115200 as console bnaud rate shows too much garbage on ipmiconsole as soon as the TFTP boot image is read. Before that, the BIOS display part is ok.
How to get rid of garbage on console when using 115200 baud rate?
