Go to main content

Oracle SuperCluster Configuration Backup Utility - Recovery Guide

Exit Print View

Updated: June 2020
 
 

Restore an IB Switch

A typical case where an IB switch needs to be restored is when it has been physically replaced. Oracle SuperCluster comes with three IB switches.

This task assumes these conditions:

  • At least one IB switch is still up and runs as a master switch. See Step 1.

  • The IB switch you plan to restore must be reachable over the management network.

  • The ZFS storage appliance must be up and running.

Additional information is available in these MOS notes:

  • How To Back Up and Restore Switch Settings for Sun Datacenter InfiniBand Switch 36 & Gateway Switch (Doc ID 1341944.1)

  • How To Back Up and Restore Switch Settings for Sun Network QDR InfiniBand Gateway Switch (Doc ID 1383752.1)

  1. Check if an IB switch runs as a master by connecting to it and running getmaster.

    Ensure that the state is MASTER. In situations where no master switch is available, contact Oracle Support.

    [root@switch-hostame ~]# getmaster
    Local SM enabled and running, state  MASTER
  2. Obtain and unzip the IB switch backup file.

    The IB switch is restored using the files created during the backup in osc-config-backup/ib-switches. There is one file per switch, called switch-hostname.backup.date.zip.

    Based on your recovery situation, use one of these procedures:

    • You can access the osc-config-backup data from a domain.

      Connect as root to a domain, create a directory accessible to everybody, become oscbackR user, and copy the .zip file in the directory. Finally unzip the file.

      root# mkdir /sharedPath ; chmod 777 /sharedPath
      oscbackR$ cd osc-config-backup/ib-switches
      oscbackR$ cp switch-hostname.backup.date.zip /sharedPath
      oscbackR$ exit
      root# cd /sharedPath
      root# unzip switch-hostname.backup.date.zip
      creating: switch-hostname.backup.date/
      creating: switch-hostname.backup.date/system/
      inflating: switch-hostname.backup.date/system/version.out
      inflating: switch-hostname.backup.date/system/showfruinfo.out
      inflating: switch-hostname.backup.date/system/ibstat.out
      ...
    • You cannot access the osc-config-backup data directly from a domain.

      If no domain is available to access the backup data, the .zip file can be transferred directly from the ZFS storage appliance to the switch by using SFTP. Ensure that SFTP is enabled on the ZFS storage appliance and that root SFTP is permitted. Set this parameter through the BUI, by selecting SFTP in the Configuration→Services tab. Also, on the ZFS storage appliance, the SFTP server listens on port 218. This is not the default SFTP port so it must be specified in the sftp command.

      [root@switch-hostame ~]# cd /tmp/
      [root@switch-hostame ~]# sftp -o port=218
      zfssaIPAddress:/export/osc-config-backup/oscbackup/ib-switches/switch-hostname.backup.date.zip
      Connecting to zfssaIPAddress...
          Password: 
      Fetching /export/osc-config-backup/oscbackup/ib-switches/switch-hostname.backup.date.zip to 
      ./switch-hostname.backup.date.zip
      /export/osc-config-backup/oscbackup/ib-switches/switch-hostname.backup.date.zip  100%  179KB 179.2KB/s   00:00
  3. Restore configuration files on the IB switch.

    These files can be restored to the IB switch using scp from a domain where the backup files are available.

    • /etc/sysconfig/network-scripts/ifcfg-eth0 (network configuration)

    • /etc/resolv.conf (DNS information)

    • /etc/ntp.conf (NTP information)

    • /etc/opensm/opensm.conf (openSM settings)

    For example, to locate the ifcfg-eth0 file with the find command and then copy it to the switch, type.

    root# cd /sharedPath/switch-hostname.backup.date
    root# find . -name ifcfg-eth0
          ./eis/ifcfg-eth0
    root# cd ./eis
    root# scp hosts root@switch-hostname:/etc/sysconfig/network-scripts/ifcfg-eth0
          hosts 100%
    |*********************************************************|
    238 00:00
  4. Update the/etc/sysconfig/network file.

    Restore the host name of the switch on the management network by adding a HOSTNAME=switch-hostname line to the /etc/sysconfig/network file, if not already present.

    After the restore, the file should look like this:

    [root@switch-hostname ~]# cat /etc/sysconfig/network
    NETWORKING=yes
    NETWORKING_IPV6=no
    HOSTNAME=switch-hostname.domainname
  5. Update the /etc/hosts file.

    Add the IP address and the host name of the IB switch on the management network to the file, if not already present. After the restore, the file should look like this:

    [root@switch-hostname ~]# cat /etc/hosts
    # Do not remove the following line, or various programs
    # that require network functionality will fail.
      127.0.0.1  localhost.localdomain localhost
      ::1  localhost6.localdomain6 localhost6
      w.x.y.z switch-hostname.domainname switch-hostname

    ...where w.x.y.z is the IB switch IP address.

  6. Update the /etc/localtime file.

    Check that the time zone specified in the /etc/localtime file is correct.

    If the time zone needs to be modified and the switch runs 1.1.3-2 or later firmware, the change must be applied through Oracle ILOM. Enter the IB switch IP address in a browser, connect as ilom-admin, and follow the Configuration/Timezone tabs.

    If you don't know which firmware version the switch is running, try to connect to Oracle ILOM, and if Oracle ILOM is present, use it.

  7. Restart the network services.
    [root@switch-hostname ~]# service network restart
    [root@switch-hostname ~]# service ntpd restart
  8. If needed, restart the OpenSM daemon.

    Check the opensm.log file in osc-config-backup/ib-switches/tmp/switch-hostname.backup.date/system to see if the OpenSM daemon was running and if so, restart it.

    [root@switch-hostname ~]# disablesm
    [root@switch-hostname ~]# enablesm
  9. Disable root SFTP and SFTP on the ZFS storage appliance.