Description

Under certain circumstances with heavy I/O load, Sun Virtual Library Extension (VLE) systems running version 1.3.12 or earlier versions of the code may cause the system to hang and panic. As a result of this specific panic, ZFS(1M) may mark the vsbpool as corrupted, leaving a number of Virtual Multi-Volume Cartridges (VMVC) and Virtual Tape Volumes (VTV) in a corrupted state.

Occurrence

This issue can occur on the following platforms:

• Sun Virtual Library Extension (VLE) systems running 1.3.12 or earlier

Note: All/any systems that run Sun Virtual Library Extension (VLE) 1.3.12 or earlier (which runs on Solaris 11.0) are potentially affected by this issue.

Symptoms

When the Serial Attached SCSI (SAS) driver encounters an inbound I/O queue stall during heavy I/O, all read/write operations to the disks can hang and the VLE can experience a system panic. As a result of the panic, VLE may end up with data check errors on VMVC/VTV data. 

Workaround

If this issue has already occurred, the data recovery procedure may require that the VLE or VLEs encountering the issue may need to be off-lined. To complete the data recovery, automatic recovery and disk-rebuilding procedures may need to be invoked multiple times. As a result, the time period that VLE needs to be off-lined will depend on the size and depth of the errors; it may take from a few days to a week or more. Please Contact Oracle Support to work with the local Field Engineer and support team to recover the data.

Note: While most of data can be recovered, it may not always be possible. There may be cases where some of the VTV data cannot be recovered, or in very rare circumstances all of the VLE data cannot be accessed/recovered, requiring some or all of the VLE data to be re-migrated.

This issue is addressed in the following release:

• Sun Virtual Library Extension (VLE) 1.4.0 or later

Please reference MOS Knowledge <Document 1483386.1> for VLE Code Release information.

Note: VLE 1.4.0 has fixes for this specific integrity problem, both in terms of driver and OS updates. In addition, there a number of VLE and OS improvements and bug fixes in the VLE 1.4.0 code. It is strongly strongly recommended to upgrade to VLE 1.4.0 to prevent possible data accessibility or loss issues.

History

22-Jan-2014: Document released; status Resolved

The expected downtime to upgrade from 1.3.12 to 1.4.0 is 3 to 4 hours,
but more time will be required if upgrading from an older version of VLE.

Questions regarding this document should be addressed to
sunalertpublication_us_grp@oracle.com and copy the
responsible engineer/submitter listed below.

Internal Contributor/Submitter: kadir.acharya@oracle.com
Internal Eng Responsible Engineer: kadir.acharya@oracle.com
Internal Eng Business Unit Group: Storage, Tape Virtual Library
Internal Associated SRs:
Internal Resolution Patches: VEL 1.3.12 A2
Reviewer (approved by): Cliff.Lawson@Oracle.Com

References

This solution has no attachment