Troubleshooting, Monitoring, and Reporting
Domain 6 - Troubleshooting, Monitoring, and Reporting
This domain tests disciplined diagnosis: read the right log, identify the failing component, and use the right report to prove service levels.
Core rules
- Start from the job. The job details show the phase that failed - scan, data transfer, index update, media mount or copy - and the component that reported it. Only then open the log for that component: CommServe, MediaAgent, client or library.
- Common pending states are informative. Waiting for resources usually means no free streams, no available drive, or an offline MediaAgent. Waiting for media usually means no spare media in the scratch pool or the library needs an inventory.
- A completed-with-errors job still wrote data; a killed job may leave partial data that later restores may depend on. Do not delete jobs casually to clean up a console view.
- Monitoring uses alerts on job failure, SLA misses and capacity thresholds, plus the dashboard for active jobs and offline clients.
- Reports answer specific questions: job summary for outcomes, client backup reports for coverage, SLA reports for compliance with agreed windows, capacity and license reports for planning.
- Performance problems are usually resource starvation: insufficient streams, multiplexing tuned too high for the drive count, or a saturated network path between client and MediaAgent.
Worked example
Ten backups are pending and nothing is running. Open one job and note it is waiting for resources on a specific storage policy. Check the library: one drive is offline and the others are occupied by a long synthetic full. The fix is restoring the drive and either raising streams or moving the synthetic full outside the backup window. The log review confirms the MediaAgent repeatedly requested a drive and was refused, ruling out client-side causes.
Traps that cost marks
- Reading the client log first when the failure was reported by the CommServe or MediaAgent.
- Treating "pending" as a fault when it is waiting for a resource you can identify.
- Deleting failed jobs before capturing logs, removing the evidence needed for a case.
- Reporting success from job status alone when SLA reports show windows were missed.