Writer Errors and Vetoes

A writer can fail for numerous programmatic reasons. When this happens, it should veto the ongoing backup, restore, or shadow copy operation by calling the CVssWriter::SetWriterFailure method in one of its handler methods (for example, CVssWriter::OnFreeze or CVssWriter::OnPreRestore) and returning TRUE. It can also optionally set an error message string in response to a failure condition in certain handler methods with the IVssComponentEx::SetPrepareForBackupFailureMsg, IVssComponentEx::SetPostSnapshotFailureMsg, IVssComponent::SetPreRestoreFailureMsg, and IVssComponent::SetPostRestoreFailureMsg methods. The requester can accept the veto or continue with the backup, ignoring the veto.

A requester should check the writer status (using IVssBackupComponents::GatherWriterStatus and IVssBackupComponents::GetWriterStatus) following each event it generates.

In some cases, error messages can be retrieved from these failures (using the IVssComponentEx::GetPrepareForBackupFailureMsg, IVssComponent::GetPreRestoreFailureMsg, IVssComponentEx::GetPostSnapshotFailureMsg, and IVssComponent::GetPostRestoreFailureMsg methods), or a writer may choose to set metadata (using IVssComponent::SetRestoreMetadata and IVssComponent::SetBackupMetadata with error state information). For example code that shows how to view such error messages, see IVssComponentEx::GetPrepareForBackupFailureMsg.

Depending on the error state, a requester or its operator could restart the backup and shadow copy with any necessary modification to the state of the backup job or system.

For example, suppose GetWriterStatus returned the following:

  • VSS_E_WRITERERROR_INCONSISTENTSNAPSHOT suggests that a requester might add additional volumes to the shadow copy

  • VSS_E_WRITERERROR_RETRYABLE indicates that retrying without reconfiguration might work. If the writer continues to return the error after several retries, try restarting the service that hosts the writer. The following writers are hosted in the VSS service: registry writer, COM+ class registration database writer, shadow copy optimization writer, and Automated System Recovery (ASR) writer. If the writer belongs to an application that hosts the writer in its own process, try restarting the application.

    Windows Server 2003 and Windows XP: The following writers are hosted in the VSS service: registry writer, COM+ class registration database writer, application event log writer, and Microsoft SQL Server 2000 Desktop Engine (MSDE) writer.

  • VSS_E_WRITER_STATUS_NOT_AVAILABLE indicates that a writer may have reached the maximum number of available backup and restore sessions, and retrying might work when the system is less busy.

  • VSS_E_WRITERERROR_OUTOFRESOURCES or VSS_E_WRITERERROR_TIMEOUT might suggest that system load be reduced prior to retrying

  • VSS_E_WRITERERROR_NONRETRYABLE or VSS_E_WRITER_NOT_RESPONDING would likely indicate a writer error so severe as to preclude trying to back up its data with VSS.

Depending on which writer and which components generate them, it is not always necessary for a backup application to abort following a veto or error.

For example, a requester may decide that the intent of the shadow copy is to back up application A and the veto has been received from the writer for backup application B. In this case, it is perfectly acceptable to continue to back up application A while ignoring the veto.

The following are examples of a writer veto:

  • The writer vetoes the shadow copy creation process when it could not suspend its activities during the time the shadow copy was being created. This indicates that there is a high probability that the shadow copy is not valid because a write operation has occurred during the Freeze state.
  • A backup application has requested a shadow copy of only volume C: and a writer determines that a shadow copy of C: and D: is to back up its data. In this case, the writer will veto. The backup application may examine the metadata and determine whether the writer will be ignored or the shadow copy creation process will be aborted and later restarted.