Addition of new node(s) to the cluster
Addition of new nodes to the cluster is assisted by the script /root/tools/newnode.py and auto-configuration file of SGE.
Detecting new MAC and population of dhcpd.conf
For new node detection script, all hostnames should be defined already in /etc/hosts . The new host name generation is based on the variables HostNamePrefix and HostNumberString (defined in the beginning of the script). Please modify, if needed.
Start a script with the argument "newmac":
sudo /root/tools/newnode.py newmac
The script will first parse the existing /var/log/syslog and ask whether to add, blacklist, or ignore each new MAC address mentioned in syslog next to DHCPDISCOVER that is not yet in the database. The database of MACs is in the files
/etc/dhcp3/cluster_hosts.list
/etc/dhcp3/macs_to_ignore.list
As soon as /var/log/syslog is processed, newnode.py will sleep for 1 second and check whether any new MACs appeared. This is a time to start a new node and wait till it will be detected. To stop the script press Ctrl-C. After each new MAC, the databases mentioned above are updated as well as a dhcpd.conf file section ( /etc/dhcp3/dhcpd_cluster_section.conf ) that describes nodes in the cluster. If you gave a wrong reply to any particular MAC, this can be solved by stopping a script, deleting a corresponding line from one of the database files and restarting a script again.
Updating list of SGE nodes
Before updating SGE list, restart dhcp daemon (to read new dhcp configuration) and reboot the nodes. Check whether you the nodes indeed booted up.
After a check, start a script with the argument "sge":
sudo /root/tools/newnode.py sge
This will process the list of nodes generated earlier ( /etc/dhcp3/cluster_hosts.list ) and add as an execution host to SGE using configuration file /opt/sge/cens_inst_template.conf as a basis.
In addition, this script will update dsh /etc/dsh/machines.list file and list of SGE groups. Note, that you have to add yourself the new hostgroups using QMON to the required queues. For this, select Queues/Modify, add hostgroup and configure it if needed. Possible adjustments: change the Sequence number and number of slots allocated to the queue. Note, the larger the sequence number - the later will be the hosts used (link)