A: Test environment:
Operating system:SLES SP1 x86_64
The SMC version is:1.4.0
Second: The operation steps are as follows
1. obtain a higher-level SMC management system with DD or clone
2. after opening the new management node, first delete the previous network Udev rules and remove all the hosts added in the SMC, and then restart the system
3. according to the actual environment to modify the hosts file, extranet IP and so on
4. Add the appropriate host (can be quickly imported through the hosts.conf file)
. Create a nodes.conf file
#touch nodes.conf
MAC Address host name IP address Startup type Unique identifier description
For example :
0050455c0392 n001192.168.4.1 Boot_mode 1 node_n001
0050455C03A2 n002192.168.4.2 Boot_mode 2 node_n002
. Select "Import Host" from the "file" menu
The Host List Import tool looks like this
650) this.width=650; "src=" http://s3.51cto.com/wyfs02/M02/57/18/wKiom1SRI1zAW-4GAAF4JkFqrkE940.jpg "title=" 11.png "alt=" Wkiom1sri1zaw-4gaaf4jkfqrke940.jpg "/>
. Select thehost list file type importing.
If you change thefile type, click Refresh to update the dialog.
. Enter Thepath for the file, want Toimport or click Browse to locate the file.
. Review thelist of the hosts to the import Andun-check any of the hosts you does not want.
Errors display foritems that cannot is imported.
. Click Importto Import the list of hosts.
. Click Close.
5. Open the SMC client and COPY to regenerate a new payload such as the name "compute-20141216-payload"
6. the "root" directory of this new payload is
/opt/sgi/sgimc/imaging/root/payloads/compute-20141216-payload
7. go to the "root" directory and make the following changes
Change the hosts file
Admin:/opt/sgi/sgimc/imaging/root/payloads/compute-20141216-payload/etc#cp/etc/hosts./
To delete an existing network Udev rule
ADMIN:/OPT/SGI/SGIMC/IMAGING/ROOT/PAYLOADS/COMPUTE-20141216-PAYLOAD/ETC/UDEV/RULES.D#RM-RF 70- Persistent-net.rules
Delete existing ssh_known_hosts files in the etc/ssh directory
ADMIN:/OPT/SGI/SGIMC/IMAGING/ROOT/PAYLOADS/COMPUTE-20141216-PAYLOAD/ETC/SSH#RM-RF ssh_known_hosts
Change NIC configuration information according to actual situation
The current NIC configuration information is as follows
Admin:/opt/sgi/sgimc/imaging/root/payloads/compute-20141216-payload/etc/sysconfig/network#ls
config if-down.d ifcfg-eth0 ifcfg-ib0 Ifcfg-lo Ifroute-lo Routes
DHCP IF-UP.D ifcfg-eth1 IFCFG-IB1 ifcfg.template Providers Scripts
If at present we only need a management network Port (eth0), then do not need to delete the network port configuration information, the configuration is completed as follows
Admin:/opt/sgi/sgimc/imaging/root/payloads/compute-20141216-payload/etc/sysconfig/network#ls
Config IF-DOWN.D ifcfg-eth0 ifcfg.template providers
DHCP if-up.d ifcfg-lo ifroute-lo scripts
8. in the SMC Management window, put the payload in the VCS Management
Double-click the payload, and in the Open window, click "Check in" on the bottom left corner.
9. within the SMC Management window, a new image named "compute-20141216-image" is generated by COPY mode
Edit the image, replace the payload with the payload we made earlier , adjust the corresponding system partition and mount information, and then click "Apply"-"Check in "
Brush (deploy) Compute nodes
One . after the compute node is deployed, set the no password access step on the Admin Management node
admin:~ #ssh-keygen// Enter key
admin:~ # CD. ssh/
ADMIN:~/.SSH # CP Id_rsa.pub Authorized_keys
Admin:~/.ssh #chmod Authorized_keys
First Use ssh to access yourself in order to add the admin to the known_hosts file
admin:~ # SSH Admin
Note: If this error message occurs when you perform the above steps, the agent admittedfailure to signusing the key, and you are prompted to enter the password, you need to execute the following command
admin:~# Ssh-add
Copy the /root/.ssh directory to the /root/ directory of all COMPUTE nodes
admin:~ # scp-r/root/.ssh/ compute node host name :/root/
Note: The first time is password-based authentication, so enter the password
Finally, copy the /root/.ssh/known_hosts file to the /root/.ssh/ directory of all COMPUTE nodes .
admin:~ # scp/root/.ssh/known_hosts compute Node host name :/root/.ssh/
Other relevant configurations as required
This article is from the "Never Stop" blog, please be sure to keep this source http://wjcaiyf.blog.51cto.com/7105309/1591039
Work Notes 1:SMC Management node upgrade steps