Troubleshooting
Problem
Data read may be corrupted (zeros instead of actual data) on some Linux NFS clients. Data residing on IBM Scale Out Network Attached Storage (SONAS) is not corrupted.
Symptom
Data presented to applications may contain zeroes instead of the actual data.
SONAS may present NFS "short reads" to clients (i.e., the data sent in the response is less than requested). Under certain circumstances where NFS "short reads" may occur, the NFS client may present zeroes instead of the real data as part of the file content to the application.
Environment
Reading data from SONAS that needs to be recalled from Tivoli Storage Manager HSM may lead to "short reads," which seems to trigger this NFS client issue.
Not all "short reads" trigger this issue. If there are any "short reads" with NFS clients that have not addressed this issue, then there is a potential for data corruption on the client.
Note: Data is not corrupted within SONAS.
Some examples of NFS clients known to be affected (not a full list):
- RHEL V6.2 and below.
- CentOS V6.2 and below.
- SLES 11 and SLES 11 SP1.
Diagnosing The Problem
Information of the code problem reported by IBM is documented in the 2 defect tracker bugs: 81182 and 90411 from IBM linux bugzilla tracking system.
Note: For additional information contact RedHat and refer to the bug 817738.
Resolving The Problem
Contact your Linux distributor and apply patches to the defects noted in above bugzilla references or upgrade to a release containing the fix.
Some examples of NFS clients known to contain the fix (not a full list):
- SLES 11 SP2 and SP3
- RHEL 6.3
Was this topic helpful?
Document Information
Modified date:
17 June 2018
UID
ssg1S1004310