IBM Books

Installation and Migration Guide


|Deleting an adapter from a node

| |

|This procedure can be used to delete an adapter from a node.

|Step 1: Archive the SDR

|

|To save the current SDR attributes, run the command:

|SDRArchive

|This command saves SDR information for all system partitions.

|Step 2: Disable nodes from the switch

| |

|If you do not have a switch in your SP system, or if you have already |logically removed the SP Switch from your nodes, or if the node you are |deleting the adapter from does not contain a switch, skip this step.

|If you want to bring the switch down for all nodes, issue the |Equiesce command for each system partition. If you use the |Equiesce command, you must later restart the switch by issuing the |Estart command. Issue the Estart command prior to Step 8: Verify installation settings.

|The Switch Admin Daemon will issue the Estart command under |certain circumstances, thus re-establishing switch recovery and primary node |takeover. To see if you are using the Switch Admin Daemon, issue the |following command:

|lssrc -a | grep swtadmd

|or

|lssrc -a | grep swtadmd2

|where swtadmd is used for an SP Switch and swtadmd2 for |an SP Switch2.

|If the response returned shows that the subsystem is active, you may want |to turn off the Switch Admin Daemon before issuing the Equiesce |command. To turn off the daemon, issue the following command:

|stopsrc -s swtadmd

|or

|stopsrc -s swtadmd2

|where swtadmd is used for an SP Switch and swtadmd2 for |an SP Switch2.

|If you are working with only a few nodes, you can use the Efence |command to disable these nodes from the switch (if appropriate, first reassign |the primary node or primary backup node). To determine if one of the |nodes to be fenced is a primary or primary backup node, issue the |Eprimary command. If you need to reassign the primary or |primary backup node, issue the Eprimary command with appropriate |options. Then issue the Estart command to make your choices |take effect. You must then issue the Efence command to disable |the nodes from the switch.

|Efence -G node_number node_number ... node_number

|Step 3: Reconfigure subsystems

|

|If an IBM Virtual Shared Disk has been configured to use the adapters being |removed, the IBM Virtual Shared Disk configuration must be modified. |Refer to PSSP: Managing Shared Disks for more information on |IBM Virtual Shared Disk configuration.

|If the adapter being removed is defined to LoadLeveler in an adapter stanza |in the LoadLeveler administration file, you will need to modify your |LoadLeveler configuration. Refer to IBM LoadLeveler for AIX |5L: Using and Administering for more information.

|Step 4: Shutdown the node

|

|The node must be powered off before the adapter can be deleted. Stop |the applications, then use SP Perspectives or the cshutdown command |to shut down the node. For example, to shut down nodes 1 and 2:

|cshutdown -G -N 1 2

|Step 5: Update DCE information for a node (required for DCE)

|

|This step is only required for DCE. If you ever plan to reuse the |host name or IP address of the adapter being removed for another adapter in |another machine, this step must be performed. If you do not plan to |reuse the host name or IP address in another machine, this step is |optional.

|The following procedure assumes that the switch adapter is being removed |and that the host names for the adapters follow a standard naming |convention. You will need to tailor this procedure for your particular |adapter and naming conventions.

|If DCE is not configured in your SP system, this step can be |skipped.

|When DCE is installed on an SP system with a switch adapter, DCE sees the |switch adapter as just another network adapter, and adds DCE principal and |account entries for the switch adapter as it does for any other network |adapter.

|This step will delete those entries in the DCE registry database and in the |DCE machine self-host (machine) key file on the node for which switch adapters |were removed. Specifically, the host/css_adapter_name and |ftp/css_adapter_name entries will be deleted.

|If DCE switch entries were not created, or not created in both the DCE |registry and in the DCE self-host key files on each host, one or both of the |removal commands will fail.

|If your site uses a common naming convention for switch adapter hostnames, |you can determine if host/ and ftp/ entries for switch |adapters exist in your DCE registry by using a combination of DCE and AIX |commands.

|For example, the following command string searches for the occurrence of |the string sn in all of the principal names in the DCE |database. sn is part of the naming convention that |identifies switch adapter host names:

|dcecp -c principal catalog | grep "sn"

|The DCE principal catalog does not require any special privileges.

|Using the switch adapter hostname naming convention in the previous |example, the same type of check can be performed against the DCE self-host key |files on each node.

|Note:
This check requires that the DCE command run as root on the target |nodes. |
|dsh -avG "/usr/bin/dcecp -c keytab show self | grep sn" | dshbak -c
|Note:
Deleting switch host/ and ftp/ entries from the DCE |registry and the DCE self-host key file on each node is optional. The |existence of these entries in both or either location does not impact DCE |functionality, nor does it create a security exposure, unless the principals |associated with these entries were added to DCE access control lists in your |DCE cell. |

|This procedure deletes entries from the DCE registry. The following |examples assume that switch adapter host names must contain |sn. |

  1. |Login to the control workstation and then DCE login as a DCE cell |administrator.
    |Note:
    This step may be performed from another host configured with a DCE client in |the same cell as the nodes DCE client. |
  2. |Delete the switch-related entries.

    |Delete the accounts first, then the principles, using either DCE |dcecp commands or through DCE's SMIT interface.

    |An example is:

    |dcecp -c account delete host/c186sn01.ppd.pok.ibm.com
    |dcecp -c account delete ftp/c186sn01.ppd.pok.ibm.com
    |dcecp -c principal delete host/c186sn01.ppd.pok.ibm.com
    |dcecp -c principal delete ftp/c186sn01.ppd.pok.ibm.com
    |

|The following examples assume that switch adapter hostnames must contain |sn. The DCE commands to follow require that DCE be running |on the node, that the invoker of the commands have the DCE self-host (machine) |principal credentials, and that the commands be run on the node.

|If you have the self-host principal credentials, the output of a |klist command will contain a line starting with Ticket cache: |and a value of:

|/opt/dcelocal/var/security/creds/dcecred_ffffffff

|The following examples show the sequence of steps required when a root user |logs into a node: |

  1. |Login to each node as a root user, but you may not need to issue a DCE |login on the node because root has default access to the self-host principal |credentials.
  2. |Obtain the complete name of the self-host key file |object.
    |dcecp -c keytab cat | grep self
    |/.../c186dcecell/hosts/c186n01.ppd.pok.ibm.com/config/keytab/self
    |Note:
    The unique portion of the name begins after the |.../keytab/ string. The unique portion of the |keytab object is a valid shorthand form of the object. |
  3. |Display the switch entries in the self keytab object (key |file).
    |dcecp -c keytab show self | grep sn
    |{/.../c186dcecell/host/c186sn01.ppd.pok.ibm.com des 1}
    |{/.../c186dcecell/host/c186sn01.ppd.pok.ibm.com des 2}
    |{/.../c186dcecell/ftp/c186sn01.ppd.pok.ibm.com des 1}
    |{/.../c186dcecell/ftp/c186sn01.ppd.pok.ibm.com des 2}
  4. |Delete the switch entries using the data from Step 2 and Step 3.

    |You do not need to issue a delete (remove) command for each |version of the ftp/ and host/ entries that exist. |However, you must delete both the ftp/ and host/ |entries.

    |dcecp -c keytab remove self -member ftp/c186sn01.ppd.pok.ibm.com
    |dcecp -c keytab show self | grep sn
    |{/.../c186dcecell/host/c186sn01.ppd.pok.ibm.com des 1}
    |{/.../c186dcecell/host/c186sn01.ppd.pok.ibm.com des 2}
    |dcecp -c keytab remove self -member host/c186sn01.ppd.pok.ibm.com
    |

|This completes the switch removal process.

|Step 6: Delete switch adapters that connect nodes to the switch planes

| |

|Perform this step to delete switch adapters that connect the node from each |switch plane. You must delete all switch adapters at this point to |disconnect the node from every switch plane in the system.

|spdeladap 5 1 0 css0

|To validate the successful deletion of the adapter objects, issue:

|splstdata -s

|Step 7: Set node to customize

|

|Each node that has an adapter being deleted must be customized. Use |the spbootins commands to update the boot/install server's |response to the bootp request from the nodes. If you are |customizing nodes in more than one system partition, you must issue these |commands in each system partition. For a complete description of the |flags associated with these commands, refer to PSSP: Command and |Technical Reference.

|For example, to set the bootp response to customize for |nodes 1 and 2, issue the following command:

|spbootins -s no -r customize -l 1,2

|Step 8: Verify installation settings

|

|Make sure that the SDR has the appropriate values specified for each of the |nodes being deleted. Issue the following command to display the |values:

|splstdata -G -b

|The expected response is that nodes with adapters being deleted should be |set to customize. All other attributes should contain the |values last used to customize your nodes.

|Step 9: Run setup_server to configure the changes

|

|The setup_server command must be run to properly set up NIM on the |control workstation by issuing the following command:

|setup_server 2>&1 | tee /tmp/setup_server.out

|The output will be saved in a log file called |setup_server.out.

|If you have a node defined as a boot/install server you must also run |setup_server out on that node.

|dsh -w boot/install_node "/usr/lpp/ssp/bin/setup_server \
|    2>&1" | tee /tmp/setup_server.boot/install_node.out

|Step 10: Refresh RSCT subsystems

|

|The SDR has been updated to reflect the deleted adapters. You must |now refresh the RSCT subsystems on the control workstation and all nodes to |pick up these changes. Run syspar_ctrl on the control |workstation to refresh the subsystems on both the control workstation and on |the nodes:

|syspar_ctrl -r -G

|Step 11: Reboot the node

|

|Once the adapter has been physically deleted in the node, the node can be |rebooted to complete the logical configuration. If you have any |boot/install servers in your system, you must customize them before |customizing their clients. Reboot the nodes to be customized by using |the cstartup command:

|cstartup -G -N 1 2

|The node has finished customization when the LEDs become blank, and its |host_responds value is yes.

|Verify that the bootp_response has been set to disk by |issuing the following command:

|splstdata -G -b

|Step 12: Rejoin the nodes to the switch network

| |

|If you do not have a switch in your SP system, or if the node you are |deleting the adapter from does not contain a switch, skip the remainder of |this step. In SP Switch2 systems, you can have some nodes on the switch |and some nodes off of the switch.

|If you disabled all nodes from the switch using the Equiesce |command, you must now issue the Estart command in each system |partition to rejoin the nodes to the switch network. If you disabled |only a few nodes using the Efence command, you must now issue the |Eunfence command to bring those nodes back to the current switch |network.

|If you stopped the Switch Admin Daemon in Step 2: Disable nodes from the switch, start it now by issuing the command:

|startsrc -s swtadmd

|or

|startsrc -s swtadmd2

|where swtadmd is used for an SP Switch and swtadmd2 for |an SP Switch2.


[ Top of Page | Previous Page | Next Page | Table of Contents | Index ]