PowerFlex: Corruption With RMcache Enabled

Applications that are using PowerFlex volumes are reporting corruption when RMcache and zero padding are enabled after upgrading the PowerFlex code.

Hard drive with RMcache enabled 

Symptoms

Corruptions are not evident in the PowerFlex logs.

Host SDC logs report corruption-related events. The following output is from an ESXi SDC:

2022-08-07T02:27:58.992Z cpu15:5421692)SESparse: 1895: Corrupt GTE GTE [valid: 3 gCluster: 117465940656700 gResource: 2304 unused: 1] ;
2022-08-07T02:27:58.993Z cpu15:5421692)BC: 414: read from 36a81bc9-<vm_name>-000001-sesparse.vmdk (def5 24 2b62 3 b1822de0 4309 3 0 0 0 0 0 0 0) 98304 bytes failed: Corrupt RedoLog

Impact

Data integrity

When an application writes data to the cache, only that data is dropped, leaving other nonzero data in the cache.

RMcache data that was not dropped may be returned to the client as part of a subsequent read, resulting in corruption.

Disable RMcache.

Upgrade to a newer PowerFlex version that includes the fix.

Impacted Versions

3.6.0.4 HF1

3.6.0.5

3.5.1.7

4.0

Fixed In Version

3.6.0.4 HF2

3.6.0Next

3.5.1Next

4.0.1