Study SnowPro DEA-C02 Query History and Profiles: key concepts, common traps, and exam decision cues.
The first performance question is usually not “which warehouse size?” It is “which evidence surface should I inspect first?”
| Need | Better first instinct |
|---|---|
| broad record of query execution | query history |
| operator-level view into why a query is slow | query profile |
| execution evidence for scheduled work | task history |
| performance complaint with no evidence yet | start with the correct history surface before changing compute |
| Scope of problem | Stronger first signal |
|---|---|
| broad record of executed queries | query history |
| one query’s operator-level bottleneck | query profile |
| scheduled work behavior | task history |
The exam often puts all three in the answer set. The winning move is to classify the scope first.
| If the stem says… | Strong reading |
|---|---|
| “evaluate performance metrics” | this is an evidence-reading question |
| “slow task or failing orchestration” | task-history lane may matter |
| “query bottleneck” | query profile is often the better first signal |
Professional troubleshooting starts by picking the smallest useful signal:
If you choose the wrong layer first, you add noise before you add clarity.
| Trap | Better rule |
|---|---|
| resizing compute before reading any evidence | diagnose first |
| using the same evidence source for every issue | queries and tasks expose different signals |
| blaming warehouses when the slowdown is clearly at the query-plan level | plan evidence matters |
| Scenario clue | Stronger answer shape |
|---|---|
| “what happened to recent queries overall?” | query history |
| “why is this one query slow?” | query profile |
| “what happened to this task run?” | task history |
| “no evidence yet, only a complaint” | choose the right history surface before resizing compute |
Performance-evidence questions usually reward the narrowest diagnostic surface that matches the issue. If you need operator-level detail for one slow query, use Query Profile. If the issue is scheduled pipeline execution behavior, use Task History. The weak answer usually starts from the wrong observability surface and loses the evidence the question is clearly pointing toward.