**Capture the exact query shape.** Copy the filter, sort and projection from the slow log or app code for one timing out call. Expected: you know the fields, operators and sort order, not just the collection name.

**Confirm the scan with explain.** Run explain with executionStats for that query in the mongo shell. Success check: the winning plan stage is COLLSCAN and totalDocsExamined is close to the collection size.

**Design the index in ESR order.** Put equality fields first, then the sort field, then range fields, using the actual query shape. Expected: the index definition mirrors the query instead of being one single field guess.

**Create the index safely.** Create the index in the background / online manner your MongoDB version uses by default, during a lower traffic window for very large collections. Success check: the build completes and appears in the index list for the collection.

**Re-run explain and the real query.** Run the same explain again. Expected: winning plan is IXSCAN, docs examined drops close to documents returned, and the query finishes well under the timeout.

**Check for a competing plan cache issue.** If explain still picks COLLSCAN, clear the plan cache for that collection once or restart the app after the index is ready, then recheck. Success check: new executions use the new index.

---

## Vectle citations

- Exact version: <https://vectle.com/skills/skl_33Y08Lx1yvhXM8G3CDHpqA/versions/skv_GXGnuQUKKwgixnWGU0F4Tw>
- Canonical HTML skill page: <https://vectle.com/skills/skl_33Y08Lx1yvhXM8G3CDHpqA>
- Current Markdown URL: <https://vectle.com/skills/skl_33Y08Lx1yvhXM8G3CDHpqA/markdown>

## Selected public attribution and original notices/references

```json
{
  "author": "Fin",
  "client": null,
  "notices_and_references": []
}
```