Storwize V7000 for Lenovo Gen2 and IBM SAN Volume Controller DH8 systems using the Compression Accelerator Hardware might take volumes offline when reading from compressed volumes

Resolving The Problem

Abstract

Storwize V7000 for Lenovo Gen2 and IBM SAN Volume Controller (SVC) Model DH8 systems that are running Real-Time Compression using the Compression Accelerator hardware may take volumes offline in response to read requests. In some circumstances some data may not be recoverable.

Affected releases are: v7.3.0.0, v7.3.0.1, v7.3.0.2, v7.3.0.3, v7.3.0.4

Symptom

Data that is written to compressed volumes using the Compression Accelerator hardware (Storwize V7000 for Lenovo Model 524, or IBM SVC DH8 with feature code AH1A) may be compressed in such a way that subsequent read operations will take the volume offline. The likelihood of exposure and precise range of potentially affected data is dependent on combinations of data patterns written to these volumes and is therefore not predictable.

This issue has been addressed in the V7.3.0.6 PTF release for existing systems. Any new systems shipped with V7.3.0.5 pre-installed are already protected from this issue, and do not need to be upgraded to V7.3.0.6.

In order to determine whether any existing compressed volumes have been affected, it is necessary to install V7.3.0.6 and then read all of the data on the compressed volumes to ensure that all data can be decompressed.

Reading all of the data from a compressed volume can be achieved either using host application or by using the mirroring capabilities of the SVC or Storwize V7000. More details can be found in the following PDF document:

How to Read the Entirety of Each Volume

In the event of compressed volumes being taken offline, or compressed volumes containing medium errors, contact IBM Support for further assistance.

Storwize V7000 for Lenovo Gen1 (2076 Models 124, 324, 112, 312) and IBM SVC CG8 systems are unaffected by this issue.

Affected configuration

Cross-reference information

Solution

This issue is addressed by the V7.3.0.6 PTF release.

Clients using compression on the affected hardware platforms are strongly recommended to upgrade to V7.3.0.6 as soon as possible.

Additional information

Will there be an impact to performance with this fix?

Overall system performance is unaffected.

Is upgrading from my current level of V7.3.0.x to V7.3.0.6 concurrent? How long will it take to upgrade?

Upgrading to V7.3.0.6 follows the standard upgrade procedure. It is concurrent and should take no longer than a typical upgrade, which is dependent on the size of the cluster.

What is the data pattern that causes this issue?

There are a range of data patterns and internal timings of processing these patterns that may encounter this issue. It is not possible to provide details of specific patterns and timing combinations to avoid.

I am using compression with the compression accelerator hardware, will I experience any volumes going offline as a result of this issue?

If you are using compression with the compression accelerator hardware, there is a possibility that you may experience offline compressed volumes. In order to determine if any compressed volumes have been exposed to this issue, install V7.3.0.6 and follow the procedure in the attached PDF.

If I installed V7.3.0.5 last week and was running compression, why do I need to now install V7.3.0.6?

V7.3.0.5 prevents this issue from continuing to occur, however any existing compressed volumes that have already been affected are taken offline if they receive read requests to affected parts of the volume. V7.3.0.6 prevents volumes from going offline and instead return medium errors, allowing host I/O operations to continue to unaffected areas of the volumes.