IBM Support

Data loss when HSM recall to GPFS snapshot fails

Flashes (Alerts)


Abstract

When using GPFS snapshots with Tivoli Storage Manager (TSM) for Space Management (HSM), a GPFS defect can cause the loss of migrated files in the snapshot

Content

If a snapshot of a given General Parallel File System (GPFS) contains stub files for files migrated by HSM, then the subsequent removal of migrated files from the GPFS will trigger a recall of these files to the snapshot.

Due to a defect in GPFS, the removed migrated files are not successfully recalled to the GPFS snapshot. The next time HSM reconcile runs, the migrated files will be deleted from TSM server storage. If the GPFS snapshot is restored, the affected files look like HSM stubs, but are corrupted. The data can not be recalled.
This problem is addressed by GPFS APARs IV58500 and IV59300.

The following sequence of events leads to this problem:
1. One or more files are migrated to the TSM server by HSM
2. A GPFS snapshot is created for that filesystem
3. One or more of the files migrated above are removed from GPFS
4. GPFS initiates a recall of these files to the snapshot
5. The recall fails, leaving corrupted stub files in the snapshot
6. HSM reconcile runs

At this point, the stub files in the snapshot corresponding with the removed files have not been replaced by resident files. The stub files are not recallable because they no longer exist on the TSM server. No related messages are logged in the dsmerror.log or in any other TSM log.

These files are lost, unless backup copies exist.

Affected GPFS levels

  • GPFS 3.4.0.13 to 3.4.0.28
  • GPFS 3.5.0.0 to 3.5.0.17

Note:
  • HSM 6.2 and higher releases support GPFS 3.4
  • HSM 6.3 and higher releases support GPFS 3.5


Fixing GPFS levels:

GPFS
release
First fixing level
GPFS 3.4
3.4.0.29
GPFS 3.5
3.5.0.18

Workaround:
Do not run reconcile until a fixing GPFS level is applied. This will prevent migrated copies from being deleted from the TSM server storage pools.

Who to call if there are further questions about this issue
After reviewing this document and APAR IV58500 and IV59300, if there are additional questions, contact IBM GPFS Technical Support for further assistance. Be sure to say that you are calling about APARs IV58500 and IV59300.

[{"Product":{"code":"SSSR2R","label":"Tivoli Storage Manager for Space Management"},"Business Unit":{"code":"BU058","label":"IBM Infrastructure w\/TPS"},"Component":"--","Platform":[{"code":"PF002","label":"AIX"},{"code":"PF016","label":"Linux"}],"Version":"6.2;6.3;6.4;7.1","Edition":"","Line of Business":{"code":"LOB26","label":"Storage"}}]

Document Information

Modified date:
17 June 2018

UID

swg21676586