Transit morning SLAs slip and ops has both Azure Monitor metrics and Spark UI stage data. How should they investigate?

When transit morning SLAs slip and ops has both Azure Monitor metrics and Spark UI stage data, they should correlate Azure Monitor metrics with Spark UI stages to explain the morning SLA misses. Monitor shows platform and job-level signals (failures, duration, resource pressure) while Spark UI exposes stage and task skew, shuffle, and spill; together they explain whether the miss is capacity, data shape, or a specific stage regression. Correlating Azure Monitor activity logs with Spark UI stages fails because activity logs record control-plane operations and audit events, not the morning-window performance signals like failures, duration, and resource pressure that explain SLA misses. Correlating Azure Monitor alert rules with Spark UI stages fails because alert rules define thresholds and notification logic, not the observed morning-window performance signals like failures, duration, and resource pressure that explain SLA misses. Correlating Azure Monitor metrics with Spark UI env configs fails because env configs show environment and library settings, not stage-level execution detail like skew, shuffle, and spill needed to identify the slow stage. Exam caveat: DP-750 monitoring skills expect correlation of timely Azure Monitor metrics with Spark UI stages, not activity logs, alert rules, or env configs. Operational check: align the SLA miss window in Azure Monitor with the job run, open Spark UI for that run’s stages, identify the dominant stage, and document the correlated root cause before changing capacity or code.