HP OpenVMS Systems Documentation

Content starts here

Guidelines for OpenVMS Cluster Configurations


Previous Contents Index

7.5.4.2 Basic Configuration Steps: Details

Prior to configuring a tape device on Fibre Channel, the worldwide identifier (WWID) of the device must be detected and stored, along with a device name, in the text file SYS$SYSTEM:SYS$DEVICES.DAT. This is usually accomplished by using the SYSMAN command IO FIND_WWID. However, some users prefer to choose their own devices name for the tape devices, rather than using the system-generated names assigned by FIND_WWID. In that case, the user will execute the IO CREATE_WWID command instead of IO FIND_WWID. IO CREATE_WWID will be described in the next section, while this current section documents the use of IO FIND_WWID.

The IO FIND_WWID command probes all ports on the Fibre Channel and locates all tape and medium changer devices connected to an MDR or NSR. For tapes and medium changers that have not been detected by a previous IO FIND_WWID command, IO FIND_WWID assigns a device name, retrieves the WWID of the device, stores the device name and WWID data in the SYS$SYSTEM:SYS$DEVICES.DAT file, and updates memory structures.

Since the primary goal of IO FIND_WWID is to populate the SYS$DEVICES.DAT file, you need to invoke the IO FIND_WWID command only once for each device. IO FIND_WWID does not configure the $2$MGAnnnn: device for use by an application.

Once the information is stored in the file, subsequent use of the IO AUTOCONFIGURE command reads a memory-resident copy of the file and configures the tape and medium changer devices automatically, loading or connecting the device drivers as needed. The SYS$DEVICES.DAT file is read into memory during each system reboot; this action initiates the automatic configuration of tapes and medium changers on the Fibre Channel.

Note that running the IO FIND_WWID command for the first time detects all existing tape and medium changer devices on the system. If you add additional Fibre Channel tape devices to the system at a later time, you must first powercycle the MDR to update internal mapping information, and then run the IO FIND_WWID command again to append the new device information to the SYS$DEVICES.DAT file. On an NSR, edit the indexed map to update mapping information.

In an OpenVMS Cluster environment, various data structures in memory must be updated on each system when a new Fibre Channel tape device is added. To accomplish this, HP recommends that you run the SYSMAN IO FIND_WWID command on each Alpha node in the cluster. Alternatively, you can run IO FIND_WWID on one node, and then reboot the other nodes that share that same system disk, because the SYS$DEVICES.DAT file is read at boot time and causes memory structures to be correctly initialized.

In the case of multiple system disks in the cluster, ensure that all copies of the SYS$DEVICES.DAT file are kept consistent, preferably by running the IO FIND_WWID command on all nodes. Alternatively, you can run IO FIND_WWID to update just one SYS$DEVICES.DAT file, and then manually edit the remaining SYS$DEVICES.DAT files by cutting and pasting the appropriate device name and WWID records from the original file to the target files. If this second alternative is used, however, the remaining nodes must be rebooted in order for the memory-resident copy of SYS$DEVICES.DAT to be updated.

HP recommends that you refrain from copying the entire original file to another system disk. The SYS$DEVICES.DAT file is also used to define port allocation classes (PACs), and PAC entries could be transferred inadvertently to the target system.

Following is a configuration example using a TL891 tape library on a single node.

First, the SYSMAN command IO FIND_WWID displays a list of all previously undiscovered tape devices and their proposed device names.


$ MCR SYSMAN IO FIND_WWID

%SYSMAN-I-OUTPUT, command execution on node SAMPLE
On port _SAMPLE$PGA0:, the following tape WWIDs and their proposed device
names have been found but not yet configured:

      [Device $2$GGA0]
      WWID=04100024:"DEC     TL800    (C) DEC3G9CCR82A017"

      [Device $2$MGA0]
      WWID=04100022:"DEC     TZ89     (C) DECCX939S2777"

      [Device $2$MGA1]
      WWID=04100022:"DEC     TZ89     (C) DECCX942S6295"

Note that the overall WWID consists of everything to the right of the equals sign. Each such WWID is unique; however, the header portion may not be unique, because the header reflects only the basic type and length of the the WWID data.

The IO FIND_WWID command automatically records the information about the new tape devices in SYS$SYSTEM:SYS$DEVICES.DAT:


$ TYPE SYS$SYSTEM:SYS$DEVICES.DAT
!
! Updated 23-OCT-2000 14:17:41.85:  DEC TL800
!
[Device $2$GGA0]
WWID=04100024:"DEC     TL800    (C) DEC3G9CCR82A017"
!
!
! Updated 23-OCT-2000 14:17:41.93:  DEC TZ89
!
[Device $2$MGA0]
WWID=04100022:"DEC     TZ89     (C) DECCX939S2777"
!
!
! Updated 23-OCT-2000 14:17:42.01:  DEC TZ89
!
[Device $2$MGA1]
WWID=04100022:"DEC     TZ89     (C) DECCX942S6295"
!

Next, the SYSMAN command IO AUTOCONFIGURE configures the tape device.


$ MCR SYSMAN IO AUTOCONFIGURE/LOG

%SYSMAN-I-OUTPUT, command execution on node SAMPLE
%IOGEN-I-PREFIX, searching for ICBM with prefix SYS$
%IOGEN-I-PREFIX, searching for ICBM with prefix DECW$
%IOGEN-I-SCSIPOLL, scanning for devices through SCSI port PKA0
%IOGEN-I-SCSIPOLL, scanning for devices through SCSI port PKB0
%IOGEN-I-FIBREPOLL, scanning for devices through FIBRE port PGA0
%IOGEN-I-CONFIGURED, configured device GGA0
%IOGEN-I-CONFIGURED, configured device MGA0
%IOGEN-I-CONFIGURED, configured device MGA1

Finally, the SHOW DEVICE/FULL command displays the WWID of the tape device.


$ SHOW DEVICE/FULL $2$MG

Magtape $2$MGA0: (SAMPLE), device type TZ89, is online, file-oriented device,
   available to cluster, error logging is enabled, controller supports
    compaction (compaction  disabled), device supports fastskip.

   Error count                    0    Operations completed                  0
   Owner process                 ""    Owner UIC                      [SYSTEM]
   Owner process ID        00000000    Dev Prot            S:RWPL,O:RWPL,G:R,W
   Reference count                0    Default buffer size                2048
   WWID   04100022:"DEC     TZ89     (C) DECCX939S2777"
   Density                  default    Format                        Normal-11
   Allocation class               2

   Volume status:  no-unload on dismount, position lost, odd parity.

Magtape $2$MGA1: (SAMPLE), device type TZ89, is online, file-oriented device,
  available to cluster, error logging is enabled, controller supports
    compaction (compaction  disabled), device supports fastskip.

   Error count                    0    Operations completed                  0
   Owner process                 ""    Owner UIC                      [SYSTEM]
   Owner process ID        00000000    Dev Prot            S:RWPL,O:RWPL,G:R,W
   Reference count                0    Default buffer size                2048
   WWID   04100022:"DEC     TZ89     (C) DECCX942S6295"
   Density                  default    Format                        Normal-11
   Allocation class               2

   Volume status:  no-unload on dismount, position lost, odd parity.

The F$GETDVI lexical function also retrieves the displayable WWID:


$ write sys$output f$getdvi("$2$MGA0","WWID")
04100022:"DEC     TZ89     (C) DECCX939S2777"

Once the device is named and configured, you can use the device in the same way that you use parallel SCSI tapes with DCL commands such as INITIALIZE, MOUNT, BACKUP, and COPY. Refer to the installation guide for individual tape layered products for details on product-specific support of Fibre Channel tapes.

Note that while medium changers on Fibre Channel are autoconfigured, the medium changers on parallel SCSI continue to require the IO CONNECT command to load the device driver. It is impossible to manually connect a Fibre Channel medium changer by the SYSMAN IO CONNECT command because the device name does not imply the device's physical location, as it does in parallel SCSI.

7.5.4.3 Creating User-Specified Device Names

If you prefer to choose specific names for the tape devices instead of using the default names generated by IO FIND_WWID, you can use the IO CREATE_WWID command. For example:


SYSMAN > IO CREATE_WWID $2$MGA3/WWID=04100022:"DEC   TZ89   (C) DECCX939S2341"

The selected name must be of the form $2$GGAn for medium changers and $2$MGAn for tapes, where n is less than or equal to 9999. The name must not be in use elsewhere in the cluster. The WWID should be cut and pasted from the output of the IO LIST_WWID display. The IO CREATE_WWID command is intended only for naming new devices; it should not be used to rename existing devices. (Renaming existing devices is discussed in Section 7.5.5.)

The following configuration example uses IO CREATE_WWID to create user-specified device names for two tapes and a medium changer within an ESL library. The commands are excuted clusterwide on a 2-node cluster consisting of nodes SYSTM1 and SYSTM2. Each node has two Fibre Channel host bus adapters, PGA0 and PGB0, so multiple paths to the tape are configured.

First, the SYSMAN command IO LIST_WWID displays a list of all previously undiscovered tape devices.


Systm1> mcr sysman
SYSMAN> set env/clus
%SYSMAN-I-ENV, current command environment:
        Clusterwide on local cluster
        Username SYSTEM       will be used on nonlocal nodes

SYSMAN> io list_wwid

%SYSMAN-I-OUTPUT, command execution on node SYSTM2
On port _SYSTM2$PGA0:, the following tape WWIDs are not yet configured:

Target 8, LUN 1, HP       ESL9000 Series
WWID=0C000008:0050-8412-9DA1-0026

Target 8, LUN 2, COMPAQ   SDLT320
WWID=02000008:500E-09E0-0009-84D1

Target 8, LUN 3, COMPAQ   SDLT320
WWID=02000008:500E-09E0-0009-4E4E

On port _SYSTM2$PGB0:, the following tape WWIDs are not yet configured:

Target 6, LUN 1, HP       ESL9000 Series
WWID=0C000008:0050-8412-9DA1-0026

Target 6, LUN 2, COMPAQ   SDLT320
WWID=02000008:500E-09E0-0009-84D1

Target 6, LUN 3, COMPAQ   SDLT320
WWID=02000008:500E-09E0-0009-4E4E

%SYSMAN-I-OUTPUT, command execution on node SYSTM1
On port _SYSTM1$PGA0:, the following tape WWIDs are not yet configured:


Target 6, LUN 1, HP       ESL9000 Series
WWID=0C000008:0050-8412-9DA1-0026

Target 6, LUN 2, COMPAQ   SDLT320
WWID=02000008:500E-09E0-0009-84D1

Target 6, LUN 3, COMPAQ   SDLT320
WWID=02000008:500E-09E0-0009-4E4E


On port _SYSTM1$PGB0:, the following tape WWIDs are not yet configured:


Target 5, LUN 1, HP       ESL9000 Series
WWID=0C000008:0050-8412-9DA1-0026

Target 5, LUN 2, COMPAQ   SDLT320
WWID=02000008:500E-09E0-0009-84D1

Target 5, LUN 3, COMPAQ   SDLT320
WWID=02000008:500E-09E0-0009-4E4E


%SYSMAN-I-NODERR, error returned from node SYSTM1
-SYSTEM-W-NOMORENODE, no more nodes
SYSMAN>

The previous NOMORENODE error is normal, because the command has completed on all existing nodes. Next, still in the same SYSMAN session, the user executes IO CREATE_WWID to choose device names $2$GGA40, $2$MGA40, $2$MGA41.



SYSMAN> io create_wwid $2$GGA40/WWID=0C000008:0050-8412-9DA1-0026
%SYSMAN-I-NODERR, error returned from node SYSTM1
-SYSTEM-W-NOMORENODE, no more nodes
SYSMAN> io create_wwid $2$mga40/WWID=02000008:500E-09E0-0009-84D1
%SYSMAN-I-NODERR, error returned from node SYSTM1
-SYSTEM-W-NOMORENODE, no more nodes
SYSMAN> io create_wwid $2$mga41/WWID=02000008:500E-09E0-0009-4E4E
%SYSMAN-I-NODERR, error returned from node SYSTM1
-SYSTEM-W-NOMORENODE, no more nodes
SYSMAN>


The user now executes IO AUTOCONFIGURE to configure the devices. Note that both the PGA path and the PGB path are configured for each node.




SYSMAN> io auto/lo

%SYSMAN-I-OUTPUT, command execution on node SYSTM2
%IOGEN-I-PREFIX, searching for ICBM with prefix SYS$
%IOGEN-I-PREFIX, searching for ICBM with prefix DECW$
%IOGEN-I-FIBREPOLL, scanning for devices through FIBRE port PGA0
%IOGEN-I-CONFIGURED, configured device GGA40
%IOGEN-I-CONFIGURED, configured device MGA40
%IOGEN-I-CONFIGURED, configured device MGA41
%IOGEN-I-FIBREPOLL, scanning for devices through FIBRE port PGB0
%IOGEN-I-CONFIGURED, configured device GGA40
%IOGEN-I-CONFIGURED, configured device MGA40
%IOGEN-I-CONFIGURED, configured device MGA41

%SYSMAN-I-OUTPUT, command execution on node SYSTM1
%IOGEN-I-PREFIX, searching for ICBM with prefix SYS$
%IOGEN-I-PREFIX, searching for ICBM with prefix DECW$
%IOGEN-I-FIBREPOLL, scanning for devices through FIBRE port PGA0
%IOGEN-I-CONFIGURED, configured device GGA40
%IOGEN-I-CONFIGURED, configured device MGA40
%IOGEN-I-CONFIGURED, configured device MGA41
%IOGEN-I-FIBREPOLL, scanning for devices through FIBRE port PGB0
%IOGEN-I-CONFIGURED, configured device GGA40
%IOGEN-I-CONFIGURED, configured device MGA40
%IOGEN-I-CONFIGURED, configured device MGA41
%SYSMAN-I-NODERR, error returned from node SYSTM1
-SYSTEM-W-NOMORENODE, no more nodes
SYSMAN> exit
Systm1>
Systm1> sho dev/fu $2$GG

Device $2$GGA40:, device type Generic SCSI device, is online, shareable, device
    has multiple I/O paths.

    Error count                    0    Operations completed                  0
    Owner process                 ""    Owner UIC                      [SYSTEM]
    Owner process ID        00000000    Dev Prot    S:RWPL,O:RWPL,G:RWPL,W:RWPL
    Reference count                0    Default buffer size                   0
    WWID   0C000008:0050-8412-9DA1-0026

  I/O paths to device              2
  Path PGA0.1000-00E0-0242-86ED (SYSTM1), primary path, current path.
    Error count                    0    Operations completed                  0
  Path PGB0.1000-00E0-0222-86ED (SYSTM1).
    Error count                    0    Operations completed                  0

Systm1> sho dev/fu $2$MG

Magtape $2$MGA40: (SYSTM1), device type COMPAQ SDLT320, is online, file-oriented
    device, available to cluster, device has multiple I/O paths, error logging
    is enabled, device supports fastskip (per_io).

    Error count                    0    Operations completed                  2
    Owner process                 ""    Owner UIC                      [SYSTEM]
    Owner process ID        00000000    Dev Prot            S:RWPL,O:RWPL,G:R,W
    Reference count                0    Default buffer size                2048
    WWID   02000008:500E-09E0-0009-84D1
    Density                  default    Format                        Normal-11
    Host name               "SYSTM1"    Host type, avail AlphaServer DS10 466 MHz, yes
    Alternate host name     "SYSTM2"    Alt. type, avail AlphaServer DS10 466 MHz,  no
    Allocation class               2

  Volume status:  no-unload on dismount, position lost, odd parity.

  I/O paths to device              2
  Path PGA0.1000-00E0-0242-86ED (SYSTM1), primary path, current path.
    Error count                    0    Operations completed                  1
  Path PGB0.1000-00E0-0222-86ED (SYSTM1).
    Error count                    0    Operations completed                  1


Magtape $2$MGA41: (SYSTM1), device type COMPAQ SDLT320, is online, file-oriented
    device, available to cluster, device has multiple I/O paths, error logging
    is enabled, device supports fastskip (per_io).

    Error count                    0    Operations completed                  0
    Owner process                 ""    Owner UIC                      [SYSTEM]
    Owner process ID        00000000    Dev Prot            S:RWPL,O:RWPL,G:R,W
    Reference count                0    Default buffer size                2048
    WWID   02000008:500E-09E0-0009-4E4E
    Density                  default    Format                        Normal-11
    Host name               "SYSTM1"    Host type, avail AlphaServer DS10 466 MHz, yes
    Alternate host name     "SYSTM2"    Alt. type, avail AlphaServer DS10 466 MHz,  no
    Allocation class               2

  Volume status:  no-unload on dismount, position lost, odd parity.

  I/O paths to device              2
  Path PGA0.1000-00E0-0242-86ED (SYSTM1), primary path, current path.
    Error count                    0    Operations completed                  0
  Path PGB0.1000-00E0-0222-86ED (SYSTM1).
    Error count                    0    Operations completed                  0

Systm1>

7.5.5 Changing the Name of an Existing Fibre Channel Tape Device

Because SYS$SYSTEM:SYS$DEVICES.DAT is a text file, you can edit it but only to change the unit number of a Fibre Channel tape or medium changer device. However, as stated earlier, Fibre Channel tape and medium changer device information is stored internally by OpenVMS using clusterwide data structures, specifically clusterwide logical names. To clean up these data structures, you must do a complete cluster shutdown. A rolling reboot (leaving at least one node up during the reboot of other nodes) is inadequate to clean up the structures.

The specific steps for changing an existing device name follow:

  1. Edit the SYS$SYSTEM:SYS$DEVICES.DAT file to change the unit number of the chosen device. In the basic $2$MGAnnnn or $2$GGAnnnn format, only the nnnn portion can be edited. The maximum allowed value for nnnn is 9999 and must be decimal. Be sure to choose a unit number that is not already in use by another device for that device type.
    For example, if $2$MGA26 is already associated with the WWID of another tape, then choose a unit number other than 26; if $2$GGA4 is already associated with the WWID of another medium changer, then choose a unit number other than 4.
  2. If there are multiple system disks in the cluster, edit each of the SYS$DEVICES.DAT files in the same way.
  3. Shut down the entire cluster to clean up existing cluster data structures.
  4. Reboot the cluster. The new device names will automatically appear.

7.5.6 Moving a Physical Tape Device on Fibre Channel

When you move a tape or medium changer device without changing its name, rebooting is not required. However, you must ensure that the NSR or MDR has assigned a FC LUN to the device at its new location, and you must then run SYSMAN IO AUTOCONFIGURE to configure the new physical path to the device. For changers only, you must also manually switch the changer to the new path using the SET DEVICE/SWITCH/PATH=new_path command. The previous paths will still show up in the SHOW DEV/FULL display, but those paths will be stale and unused, with no harmful side effects; after the next reboot the stale paths will disappear.

7.5.7 Swapping Out an NSR on Fibre Channel

You can swap out an NSR without rebooting the Alpha OpenVMS system. (This capability was introduced in OpenVMS Alpha Version 7.3-1.)

After attaching the new NSR, use the Mapping submenu in the Visual Manager to populate the Indexed map on each Fibre Channel port of the NSR and reboot the NSR. An alternative way to map the new NSR is to copy the .cfg file from the previous NSR via the NSR's FTP utility.

Once the Indexed map is populated, run SYSMAN IO AUTOCONFIGURE to configure the new physical paths to the tape. For changers only, you must also manually switch the changer to the new path using the SET DEVICE/SWITCH/PATH=new_path command. The previous paths will still show up in the SHOW DEV/FULL display, but those paths will be stale and unused, with no harmful side effects; after the next reboot the stale paths will disappear.

7.5.8 Serving a Fibre Channel Tape Device

In general, all OpenVMS Alpha Version 7.2-2 or later nodes in an OpenVMS Cluster have a direct path to Fibre Channel tape devices if the nodes are connected to the same Fibre Channel fabric as the NSR (or MDR). VAX nodes and Alpha nodes running earlier versions of OpenVMS, as TMSCP clients, can be served Fibre Channel tape devices by an OpenVMS Alpha Version 7.2-2 or later node. To provide this feature, pre-Version 7.3 VAX nodes require the latest update kit for TUDRIVER.EXE, and pre-Version 7.2-2 Alpha nodes require the latest update kit for SYS$TUDRIVER.EXE.

Medium changers, whether connected to Fibre Channel or to parallel SCSI, cannot be TMSCP served.

7.5.9 Replacing a Fibre Channel Tape Device

If one tape drive must be physically replaced by another tape drive at the same FC LUN location within the MDR or NSR, update the appropriate data structures with the IO REPLACE_WWID command.

For example, you may need to replace a defective tape drive with a new drive without rebooting the cluster, and that drive may need to retain the device name of the previous tape at that location.

The replacement device should have the same SCSI target ID as the original device. Cease all activity on the device, then type the following command to update all the necessary file and memory data structures with the WWID of the new tape drive:


$ MCR SYSMAN IO REPLACE_WWID $2$MGA1

Execute this command on each Alpha node in the cluster environment. You can accomplish this with the following commands:


$ MCR SYSMAN
SYSMAN> SET ENVIRONMENT/CLUSTER
SYSMAN> IO REPLACE_WWID $2$MGA1

In some cases, this command may fail because the device name $2$MGA1 no longer exists in the SHOW DEVICE display. This happens when the system has been rebooted some time after the drive has malfunctioned. In such a case, you must specify both the device name and the WWID, as shown in the following example.

The WWID must be the WWID of the new device that resides at the same Port/Target/LUN location as the replaced device. (To determine the value of the WWID that resides at a particular Port/Target/LUN location, use the SYSMAN IO LIST_WWID command.)


$ MCR SYSMAN
SYSMAN> SET ENVIRONMENT/CLUSTER
SYSMAN> IO REPLACE_WWID $2$MGA1/WWID=02000008:500E-09E0-0009-4E44

Note

This command should not be used to rename devices or redefine WWID correlations. The specified WWID must not be associated with any other device name in the cluster.

7.5.10 Determining the Physical Location of a Fibre Channel Tape Device

Given the name of a Fibre Channel tape device, it is helpful to know how to locate the Fibre Channel tape device. To do so, follow these steps:

  1. From the system manager's account, run ANALYZE/SYSTEM.
  2. At the SDA prompt, type CLUE SCSI/SUMMARY.
  3. Search for the name of the device (for example, MGA3) in the column labeled DEVICE.
  4. Note the corresponding value in the column labeled SCSI-LUN. This SCSI LUN value is the same value used by the MDR or NSR as the FC LUN. Also note the columns labeled Port and SCSI-ID for the device; all devices associated with that same port and SCSI-ID are attached to the same physical Fibre Channel port of the same MDR or NSR.
  5. For the NSR, enter the Mapping submenu of the Visual Manager and click on Edit/View next to the Select Map box to display the Indexed Map for the current port.
    In the column labeled FC LUN, locate the value noted in step 4. Once you find the FC LUN value, note (on the same line) the corresponding values for SCSI Bus, SCSI Target ID, and SCSI LUN. This B:T:L information describes the physical location of the device within the NSR. Additional device information is available by clicking on 'Report' in the main menu of the Visual Manager.
    If an MDR is being used, at the MDR console's AMC prompt, similar information is available by typing ShowFcScsiMap .


Previous Next Contents Index