Potential overwrite of newly ingested chunks in a directory container storage pool.
Under certain circumstances, the IBM Spectrum Protect Server might overwrite newly ingested chunks in a container.
WHO IS AFFECTED:
All IBM Spectrum Protect users with server version 8.1.1.025, 22.214.171.124xx, 8.1.4.x, and 126.96.36.199xx in which client data ingest sessions or node replication sessions store data to container storage pools.
The following conditions are required for this situation to be encountered:
1) Significant amount of concurrent sessions starting and stopping very quickly.
2) Within those concurrent session workloads, a significant amount of small files must be ingested.
3) The IBM Spectrum Protect server must be experiencing performance issues during this time frame. These performance issues must also affect the IBM Spectrum Protect server transactional processing rates.
Client ingest and node replication sessions have the ability to perform operations that can cause the inability to later restore, retrieve, or repair data.
The damage occurs during ingestion of data which means the ability to recover using the PROTECT STGPOOL or REPLICATE NODE commands is not possible. The damaged file data will need to be re-ingested from the affected client machine. If this condition is met, it will typically affect only a few files per session. The affected files will be marked damaged once they are attempted for access, either by server-side processing or client-side restore/retrieve requests.
The following type of entry will be entered into the FFDC log if this issue has been encountered:
[ FFDC GENERAL_SERVER_ERROR ]: (sdutil.c:4462) Chunk Header Invalid Signature chunkId: 0 Container:
/tsm/container/10/0000000000009999.dcf Offset: 36929536
Usually, the Invalid Signature will be encountered but another failure indicator of this issue is a Chunk Header Length Mismatch. Both will be in the FFDC and from similar line numbers in sdutil.c.
Upgrade to a level of the IBM Spectrum Protect server that contains a fix for APAR IT25408.
This problem has been fixed in the 188.8.131.52 and 184.108.40.206 levels of the IBM Spectrum Protect server.
IDENTIFYING AFFECTED DATA:
You can use the REPLICATION or PROTECT commands on a source server to detect this problem. On a target server or a server where neither replication nor protect are implemented, you can use the AUDIT CONTAINER command to detect the problem.
REPAIRING DAMAGED DATA:
For a source server delete the affected files and then back up the files again. For a replication target server, delete the files from the target server and then replicate again using the FORCERECONCILE=YES option.
Spectrum Protect server
More support for:
IBM Spectrum Protect
Software version: 8.1.1.025, 220.127.116.11xx, 8.1.4.x, 18.104.22.168
Operating system(s): AIX, Linux, Windows
Reference #: 0715659
Modified date: 24 April 2019