HPlogo Mirrored Disk/iX User's Guide: HP 3000 MPE/iX Computer Systems > Chapter 4 Troubleshooting

Troubleshooting

» 

Technical documentation

Complete book in PDF
» Feedback

 » Table of Contents

 » Glossary

 » Index

This section describes troubleshooting information for the following conditions:

  • Device adapter card drive failure.

  • Disabled drive.

  • Data recovery using DISCUTIL.

  • System abort.

  • JOINMIRRSET aborted.

  • Drive fails when volume set is split.

  • System abort while volume set is being split.

  • Missing volumes during JOINMIRRSET.

If a device adapter card's drive fails

If a disk drive is directly connected to the device adapter card and that drive fails, the remaining drives connected to that device adapter card may not be available for use, depending on the type of drive failure. If power is lost to the drive, all disks chained off that drive go DISABLED, and a REPLACEMIRRVOL command is required for each drive on the chain to resume disk mirroring.

If a drive is DISABLED

A drive could be marked DISABLED simply because of a time out (over two minutes) due to loss of power to the drive, or a read/write error that is correctable by sparing (recovering defective disk sectors). Before replacing the drive, issue a REPLACEMIRRVOL command on the disabled drive, assuming that it is powered on and responding. If it again goes DISABLED, contact your support personnel.

If DISCUTIL is needed

Use the DISCUTIL utility to save and recover data from disk drives when MPE/iX is not available.

NOTE: When using DISCUTIL, you must power off one partner of each mirrored disk pair before DISCUTIL will work properly.

For more information on DISCUTIL, use the Volume Management Reference Manual (32650-90045).

If the system aborts

In the event of a system abort, use the DUMP utility to save the current state of system memory and secondary storage to tape for later analysis.

For more information on DUMP, use the System Startup, Configuration, and Shutdown Reference Manual (32650-90042).

If JOINMIRRSET is aborted

A volume join is aborted if one of the following conditions occur:

  • A source volume is missing.

  • Your negative reply when prompted for verification.

  • A source volume fails.

  • System aborts during JOINMIRRSET.

  • If SOURCE=USER is requested, and a split-volume backup is taking place.

  • If SOURCE=BACKUP is requested, and there are files opened on the user volumes.

If a source volume is missing

The join can be retried later with either volume set half as the source. You will have an opportunity to correct the error that resulted in missing volumes. If the situation cannot be rectified, the alternate volume set half can be used as the source. If neither volume set half is complete, a reload of the entire volume set is necessary.

If a source volume fails

After the JOINMIRRSET command has proceeded to the point of starting repairs, if any of the source volumes incur a disk failure before the repair completes, the entire volume set must be reloaded. This situation can be noted by the occurrence of a disk failure on a volume for which VOLUTIL did not print an "absent" message. The join cannot be retried with the other volume set half as the source volumes since they have already been used as the destination volumes.

If the system aborts during JOINMIRRSET

If a JOINMIRRSET command is interrupted by a system abort, the volume set may be partly split. The master volume is the first to be joined, and its state determines if the entire set is split or joined; therefore if the master is joined, all the members in the set are joined. The join is completed during reboot, and the repair process starts; otherwise, the volume set mounts as a split-volume set and the JOINMIRRSET command processing must be retried.

If a drive fails when volume set is split

While a volume set is split, if it fails, it is identical to a disk failure on a non-mirrored volume set. The application accessing the failed drive hangs or causes a system abort; however, the data on the failed drive can be recovered at the time of the join as long as the split-volume set still includes a complete volume set half (user or backup volumes).

If the failed drive is a backup volume, the latest copy of the data is available from the user volume. If the failed drive is a user volume and data is recovered from the backup volumes, changes made to the user volumes after the split are lost.

If the system aborts while volume set is being split

If a volume set split or join is interrupted by a system abort, the volume set may only be partially split. Upon remount, a partly split-volume set is always automatically joined.

Interrupted join

Results in a successful join.

Interrupted split

Returns the volume set to a joined state. The volume set split has to be redone after the repair (activated by the join) is complete.

If the volumes are unavailable during JOINMIRRSET

The JOINMIRRSET command matches volumes from user and backup split-volume sets that were previously mirrored partners, and starts a repair in the direction specified by the SOURCE parameter. Ideally, all user volumes and backup volumes belonging to the original volume set are mounted; however, volumes may be unavailable for a variety of reasons: disk error, power failure, user error. The possible combinations of volume states and their resulting actions are described in the following table.

Table 4-1 JOINMIRRSET Options

SOURCE=Description
SOURCE=USER (No volumes missing)

All volumes in the set can be joined. Only the messages pertaining to the join itself are printed:



 *Note: Volume MEMBER1 on ldev 34 and ldev 36 will be joined.

 *Note: Volume MEMBER2 on ldev 35 and ldev 370will be joined.
SOURCE=USER (One or more backup volumes missing)

The partnerless user volumes are automatically suspended (the equivalent of SUSPENDMIRRVOL commands). The user volumes are continuously available for use. You can later resume mirroring by adding partners through the REPLACEMIRRVOL command.



 *Warning: Backup volume MEMBER2 is absent.

 *Note: Volume MEMBER2 on ldev 37 will be suspended.

 *Note: Volume MEMBER1 on ldev 34 and ldev 36 will be joined.
SOURCE=USER (All backup volumes missing)

All user volumes are suspended.

 *Note: No backup volumes are mounted.

 *Note: Volume MEMBER1 on ldev 34 will be suspended.

 *Note: Volume MEMBER2 on ldev 37 will be suspended.
SOURCE=USER (One or more user volumes missing)

If any of the source volumes are missing, the join cannot proceed and is aborted.

 *Warning: User volume MEMBER1 is absent.

 *Error: User volumes incomplete, cannot use as repair source.
SOURCE=BACKUP (No volumes missing)

Since this repair results in changes being made to the user volumes since the split has been lost, verification is requested before the command proceeds. All users of the volume set must be logged off.

 *Verify: User volume set changes will be lost, continue [Y/N]?

 *Note: Volume MEMBER1 on ldev 34 and ldev 36 will be joined.

 *Note: Volume MEMBER2 on ldev 35 and ldev 37 will be joined.

SOURCE=BACKUP (One or more user volumes missing)

All users of the volume set must be logged off. Volumes with partners start repairing, and backup volumes with missing partners go mirror PENDING. Verification is requested.

*Warning: User volume MEMBER2 is absent.

*Verify: User volume set changes will be lost, continue [Y/N]?

*Note: Volume MEMBER1 on ldev 34 and ldev 36 will be joined.

*Note: Use the SUSPENDMIRRVOL command for volume MEMBER2 on ldev 35.
SOURCE=BACKUP (All user volumes missing)

All backup volumes become mirror PENDING.

*Note: No user volumes are mounted.

*Verify: User volume set changes will be lost, continue [Y/N]?

*Note: Use the SUSPENDMIRRVOL command for MEMBER1 on ldev 36.

*Note: Use the SUSPENDMIRRVOL command for MEMBER1 on ldev 35.
SOURCE=BACKUP (One or more backup volumes missing)

Since some of the source volumes are missing, the join cannot proceed and is aborted.

*Warning: Backup volume MEMBER1 is absent.

*Error: Backup volumes incomplete, cannot use as repair source.

 

Example: Repairing from backup volumes

This example shows how to perform a backup from the backup volumes when the user volume set is incomplete or data on the volume set is known to be corrupt.

  1. Use the DSTAT command to find out if all of the volumes mounted. The user volume LDEV 33 did not mount. Since all of the user volumes did not mount (and all of the backup volumes did mount), the backup volumes can be used as the source of the join and the repair.

       :DSTAT
    
    
    
       LDEV-TYPE   STATUS       VOLUME  (VOLUME SET - GEN) 
    
       _ _ _ _ _ _ _ _ _ _ _    _ _ _ _ _ _ _ _ _ _ _ _ _ 
    
       30- 079370   MASTER-SU     MEMBER1     (PROD_SET-0)
    
       31- 079370   MASTER-SB     MEMBER1     (PROD_SET-0)
    
       32- 079370   MEMBER-SB     MEMBER2     (PROD_SET-0)
    
    NOTE: When you use backup volumes as the source of the join and repair, the changes that were made to the user volumes during the backup are lost. For example, a file that was created after the volume set was split, does not exist on the volume set after the join.
  2. Inform users that they must be logged off the volume set before this type of join can be requested.

       : TELL @ LOGOFF NOW FOR JOIN
    
  3. Invoke VOLUTIL.

  4. Use the JOINMIRRSET command with SOURCE=BACKUP.

       :VOLUTIL
    
    
    
       Mirvutil A.00.00, (C) Hewlett-Packard Co., 1990.
    
       All Rights Reserved.
    
    
    
       volutil:JOINMIRRSET PROD_SET SOURCE=BACKUP
    
       volutil:EXIT
    

    If any user volumes are missing, the partner backup volumes are not automatically suspended by the JOINMIRRSET command; however, they are placed in the PENDING state after the join.

  5. Exit VOLUTIL.

  6. Notify users that the volume set is available.

  7. Use the DSTAT command to find out which volumes are pending.

       : TELL @ SYSTEM IS AVAILABLE NOW
    
       : DSTAT
    
    
    
       LDEV-TYPE   STATUS       VOLUME  (VOLUME SET - GEN) 
    
       _ _ _ _ _ _ _ _ _ _ _    _ _ _ _ _ _ _ _ _ _ _ _ _ 
    
       30- 079370   MASTER-MD     MEMBER1     (PROD_SET-0)
    
       31- 079370   MASTER-MD     MEMBER1     (PROD_SET-0)
    
       32- 079370  *PENDING-MD    MEMBER2     (PROD_SET-0)
    
  8. Use the VOLUTIL SHOWSET command with the MIRROR option to display the state of the volumes in the mirrored set.

       :VOLUTIL
    
       
    
       Mirvutil A.00.00, (C) Hewlett-Packard Co., 1990.
    
       All Rights Reserved.
    
    
    
       volutil:SHOWSET PROD_SET MIRROR
    
    
    
       Volume Name  Vol Status   Mirr Status  Ldev  Mirr ldev
    
       - - - - - -  - - - - - -  - - - - - - - - -  - - - - - 
    
       MEMBER1      MASTER       REPAIR-DEST  30    31   
    
       MEMBER1      MASTER       REPAIR-SRCE  31    30   
    
       MEMBER2      MEMBER       PENDING      32    *
    
  9. Since the user volume LDEV 33 was missing, its partner is placed in the PENDING state after the join. To place MEMBER2 in non-mirroring operation, use the SUSPENDMIRRVOL command for LDEV 32 or use the REPLACEMIRRVOL command for a new partner for LDEV 32.

Feedback to webmaster