Powering on the system

Learn how to power on the Cloud Pak for Data System 2.x.

About this task

The procedure applies to all 2.x systems with or without NPS installed. Some steps are only required for NPS systems, as noted.


  1. Power on e1n1 manually or from the KVM using IPMI:
    ipmitool -I lanplus -H e1n1bmc -U USERID -P PASSW0RD power on
    Note: Make sure you only power on e1n1 node. The remaining nodes must remain offline at this point.

    Wait for e1n1 to fully power up and boot NodeOS. This takes about 6 minutes.

  2. Once e1n1 is reachable, log onto e1n1 and verify that the dhcpd service is running:
    ssh e1n1
    systemctl status dhcpd
    If it is not running, issue the commands:
    systemctl start dhcpd
    systemctl start named
    ifup fab-br0:Fl
    ifup mgt-br0:Fl
  3. From e1n1, power the rest of the nodes:
    END=$(cat /etc/dhcp/static-mbond.conf | egrep ddns | egrep smm | wc -l); for (( c=1; c<=$END; c++ )); do for dev in e"$c"smm e"$c"n{1..4}bmc; do echo "[$dev]"; ipmitool -I lanplus -H $dev -U USERID -P PASSW0RD power on; done; done
  4. Verify that all nodes are reachable:
    for node in e{1..n}n{1..4}.mbond; do ping -c 3 $node; done
    Note: On NPS systems, the SPU nodes will not be reachable via mbond until you scale the host pod (pods) up. This is expected. Proceed with the following step.
  5. Verify that BMCs are reachable:
    for ip in e{1..n}n{1..4}bmc; do echo $ip; ping -c 3 $ip;done
  6. Verify that SMMs are reachable:
    for ip in e{1..x}smm; do echo $ip; ping -c 3 $ip;done

    If SMM are not pinging back:

    (The following examples are for e2smm.)

    [root@e1n1~]# ssh e2n1
    [root@e2n1~]# ipmitool raw 0x3a 0xf1 0x02
    This disables SMM port.
    [root@e2n1~]# ipmitool raw 0x3a 0xf1 0x01
    This enables SMM port.
    [root@e2n1~]# ipmitool raw 0x3a 0xf1 0x00
    This shows SMM port status. It should report 01, which means it is enabled.
    [root@e2n1~]# ping e2smm
    It might take 30 seconds before you get ping response.
  7. Start and mount GPFS file services on all control nodes. From e1n1 type:
    mmstartup -a
    mmmount all -a

    Verify that the state is active on the 3 control nodes:

    [root@e1n1 ~]# mmgetstate -aLv
    Node number    Node Name   Quorum     Nodes up    Total nodes    GPFS state    Remarks
        1            e1n1         2           3             3          active      quorum node 
        2            e1n2         2           3             3          active      quorum node 
        3            e1n3         2           3             3          active      quorum node 
    [root@e1n1 ~]
  8. Start database and the software stack. From e1n1 type:

    Wait several minutes for the services to start.

  9. Monitor the state with apstate -d until the status is as follows:
    System state is 'Ready'
    Application state is 'Ready'
    Platform management state is 'Active'
  10. NPS systems only: Scale up the NPS host pod.
    oc scale statefulset.apps/ipshost --replicas=1 -n nps-1
    Note: In the above example the NPS namespace is named nps-1. In case of multitenant deployments, the command must be run for each NPS namespace. The following command can be used to identify all NPS namespaces:
    oc get sts --all-namespaces | grep ipshost  | awk '{print $1}'