Date: Dec 06
NV Version: All
OS Version: All
Application version: N/A
Plugin version: N/A
Description: Media Request Timeout not working
Symptoms:
Jobs appear to not adhere to the Media Timeout Request value set for them. This manifests itself as jobs appearing to start writing to media after the media timeout value for them has been reached.
Imaginary scenario:
20 jobs for various clients all submitted at the same time. All jobs have a "Media Request Timeout" of 14 hours.
Physical library with 4 drives - no VDL.
Backup window from 19:00 to 9:00 the following day (14 hours)
The problem occurs because the 20 jobs are submitted in one go, a handful start writing straight away (determined by the number of drives available - say 4 as an example) and another batch (say 12) sit waiting for media. The rest get queued because there are only a finite number of jobs that can be on the media request list. These jobs do not have media requests and therefore the media timeout clock isn't yet running for these jobs.
After the "Media Request Timeout" value has been reached (14 hours in this example) any jobs that are waiting for media get automatically aborted. At this point the media request list is emptied and the jobs that have been queued obtain media requests and the media timeout clock starts for these jobs. Some of these jobs then start writing.
Ultimately the amount of data being backed up is too great for the time window and the number of drives available. The only surefire way to resolve this is to increase the number of drives (unlikely to be possible in a library) or to use a VDL to stage the backups. You can also help to minimize the risk of this situation happening by scheduling the jobs more efficiently and using job prioritization.
