If a job fails at 3 a.m., what logs or monitoring do you check in the morning to investigate the failure, and where would you look for the error?

When a job fails at 3 a.m., the first place I look is the alerting system—most teams use PagerDuty or Opsgenie to surface critical failures. I check the alert message for the job name, timestamp, and error code. Next, I open the monitoring dashboard (Grafana, CloudWatch, Datadog) to see the job’s metrics: CPU, memory, queue depth, and any spike in error rates. I then drill into the job’s log files, which are typically stored in a centralized log store such as CloudWatch Logs or an ELK stack. I search for the 3 a.m. timestamp and look for stack traces or exception messages. If the job writes to a database, I query the transaction logs or audit tables to see if partial writes occurred. Finally, I review any downstream alerts that might have been triggered by the failure. By correlating alerts, metrics, and logs, I can pinpoint the root cause—whether it’s a resource exhaustion, a network timeout, or a data quality issue—and take corrective action.