Rocks Basic Commands

This page describe various  Rocks Commands. 

Show Hosts 

·         # rocks list host 

·           

Show all Rocks Config 

·         Rocks dump   

Rocks list host interface 

·         rocks list host interface compute-0-1 

  

Set and Sync Root Password 

  

1.         Set root password for compute nodes 

1.         [root@tdub-qm01 nodes]# rocks set host sec_attr compute attr=root_pw Enter root_pw: Confirm root_pw: 

2.         Set root password for login nodes 

1.         [root@tdub-qm01 nodes]# rocks set host sec_attr login attr=root_pw   Enter root_pw: Confirm root_pw: 

3.         List the Password for compute and login nodes 

1.         [root@tdub-qm01 nodes]# rocks list host sec_attr compute ATTR    VALUE                              ENC root_pw $1$eBBzZrUI$gXyj.gJf7i1tS1.9ALtrG. crypt 

2.         [root@tdub-qm01 nodes]# rocks list host sec_attr login HOST               ATTR    VALUE                              ENC site2-prd-lndt-1-0: root_pw $1$IcKy818Z$IgzEYECPhhoxeUDKKIIkk/ crypt site2-prd-lndt-1-1: root_pw $1$IcKy818Z$IgzEYECPhhoxeUDKKIIkk/ crypt 

1.         Sync compute and login  sec_attr 

1.         [root@tdub-qm01 nodes]# rocks sync host sec_attr compute 

2.         [root@tdub-qm01 nodes]# rocks sync host sec_attr login 

  

  

Install Node 

The default naming is:  compute-x-y where x= cabinet and y=rank  

·         General Install 

Insert-ethers  

·         In a specific location 

Insert-ethers --cabinet=3 --rank=15 

·         To remove a node from the Rocks Cluster 

Insert-ethers --remove --cabinet=3 --rank=15 

·         To add a node with a specific base name 

insert-ethers --basename=site1-prd-lncn --cabinet=6 --rank=4 

·         To clear/sync head node if not installing 

o    insert-ethers --remove --cabinet=X --rank=X 

o     rocks sync config  insert-ethers --update 

o    Insert-ethers --cabinet=X --rank=X 

·         Once the insert-ether's tui is running boot the pxe boot the node off of the interface you wish to be used for management (e.g. SGE VLAN network)  

  

·         See How to Add Rocks Compute Node for step by step guide on adding compute nodes.   

·         See compute node info in database 

o    dbreport hosts 

  

Re-kickstart Rocks  

·         rocks set host boot compute-0-0 action=install 

·         ssh compute-0-0 "shutdown -r now"  

  

Other KickStart Option 

·         rocks run host tdub-prd-lncn-'%' /boot/kickstart/cluster-kickstart-pxe 

·          or ssh [node] /boot/kickstart/cluster-kickstart-pxe 

  

SGE Force Remove Node 

·         qconf -dh HOST_GO_AWAY >/dev/null  

·         qconf -ds HOST_GO_AWAY >/dev/null 

·         qconf -de HOST_GO_AWAY >/dev/null  

  

Examples 

[root@tdub-qmaster ~]# qconf -ds tdublin23 > /dev/null 

root@tdub-qmaster removed "tdublin23" from submit host list 

[root@tdub-qmaster ~]# qconf -de tdublin23 > /dev/null 

root@tdub-qmaster removed "tdublin23" from execution host list 

[root@tdub-qmaster ~]# qconf -dh tdublin23 > /dev/null 

root@tdub-qmaster removed "tdublin23" from administrative host list 

  

  

Sync Config and Compute 

# rocks sync config 

# rocks sync host network compute-0-1 

  

Cycle through all nodes  (use for testing) 

·         rocks run host 'hostname; ls -d /isilon/; '  

  

Add a Network 

·         rocks add network data subnet=10.20.64.0 netmask=255.255.253.0 mtu=1500 

·         rocks add host interface compute-4-0 eth2 ip=192.168.1.2 subnet=data name=eth2 

  

Add Jumbo Frames 

·         rocks set network mtu data mtu=9000 

  

  

Create Distro 

1.         Verify xml file is ok, and fix any issues. 

1.         xmllint -noout extend-compute.xml 

2.         # cd /export/rocks/install # rocks create distro 

3.         Test kickstart file,run this test 

1.         /export/rocks/install/sbin//kickstart.cgi -c compute-3-0 > /dev/null 

2.         Should return nothing for okay. 

  

Restart Ganglia (optional create as script?) 

·         /etc/init.d/ganglia-restore-rrds restart /etc/init.d/gmetad restart /etc/init.d/gmond restart /etc/init.d/greceptor restart rocks run host compute "/etc/init.d/gmond restart" 

  

Sets cpus 

·         Sets the number of cpus of compute-0-0 to 2  

  

DHCP Config 

/etc/sysconfig/dhcpd 

DHCPARGS="bond0" 

  

Troubleshoot Tips 

  

to debug this, you can run: 

# rocks list host profile compute-0-0 > /tmp/ks.cfg 

then examine /tmp/ks.cfg. 

look at the top of /tmp/ks.cfg and look for a line that says: 

# ./nodes/extend-compute.xml (base) 

that means your extend-compute.xml file was included. 

then search for: 

/etc/profile.d/us.sh 

and see if that section has been included in /tmp/ks.cfg. 

  

BONDING interfaces 

  

·         rocks list host interface compute-0-1 

  

SUBNET IFACE MAC IP NETMASK MODULE NAME VLAN OPTIONS 

private eth0 00:1e:4f:b0:74:ef 10.100.101.253 255.255.0.0 tg3 compute-0-1 ---- ------- 

------- eth1 00:10:18:31:74:43 ------------ ----------- tg3 ----------- ---- ------- 

  

# rocks add host bonded compute-4-0 channel=bond0 interfaces=eth0,eth1 \ 

ip=10.100.101.253 network=private 

  

dhcp range 10.20.65.101-20 

64.31 last 

64.32 

  

# rocks sync config 

# rocks sync host network compute-0-1 

  

·         rocks add host bonded compute-0-0 channel=bond0 interfaces=eth0,eth1 ip=10.100.101.254 network=private 

·         rocks add network data subnet=10.20.64.0 netmask=255.255.252.0 

  

Rocks DNS command 

·         # rocks report zones 

·         # rocks report named