When restoring directories with a large number of files (more than 100,000), the job may remain queued and not go active. The job details may show "Begin restore" in the Activity Monitor for many hours. On the master server the bprd and a bpdbm process will be observed consuming a lot of CPU cycles.
The restore job will eventually start, but delays of longer than 17 hours have been observed.
An algorithm change was made in NetBackup 7.1 that resulted in this delay when processing flat file backup of millions of files. Normally this type of backup is recommended for a FlashBackup policy which would not be affected, as opposed to a Standard or MS-Windows policy.
The formal resolution to this issue (Etrack 2373538) is included in the following release:
- NetBackup 7.6 Maintenance Update 2 (184.108.40.206)
NetBackup 220.127.116.11 is available for download via the linked Related Article.
The restore will eventually complete if it is allowed to run long enough.
The restore can also be completed by selecting smaller subdirectories (< 100,000 files each) and submitting each as an independent restore job.
If these workaround are unfeasible and upgrade to a version where this issue is resolved is also unfeasible, please contact Veritas technical support, referencing this document and Etrack 2373528 to determine if an Emergency Engineering Binary (EEB) is available for the version of NetBackup running in the production environment. Please note that if an EEB is not available, Veritas technical support will strongly encourage an upgrade of NetBackup to mitigate this issue.
NetBackup 7.1 and above are affected by this issue.