The client direct backup fails due to client data streaming patterns

In some cases, BMR and file system backup failures may occur due to the data streaming behavior of the client's file system.

During an initial backup, all data is considered new and is written to storage. However, as the backup policy continues to run, the combination of schedule settings and acceleration attributes determines how much data is new and how much is referenced.

If most of the data is referenced, or if the access pattern is unevenly distributed, it may appear that no new data is being written. This perceived lack of progress can lead to time-outs, causing the backup to terminate prematurely.

Error message:

Info bpbkar (pid=11936) done. status: 14: file write failed 
Error nbpem (pid=921441) backup of client <host name> exited with status 14 (file write failed)

Example scenario:

In a Windows BMR backup using ALL_LOCAL_DRIVES (for example, C and E), a large portion of data may be written in the first few seconds. After that, the remaining data is mostly referenced. When drive E is snapshotted, another burst of writes occurs. Over time, these "includes" accumulate, contributing to the stalled state.

Workaround:

To resolve this issue, increase the retry limit for stalled data streaming by modifying the bp.conf configuration file. Specifically, adjust the OST_CD_BUSY_RETRY_LIMIT parameter to allow more retries before the process fails.

This overrides the default limit of 500 retries, giving the backup process more time to recover from stalled states.