- Industry
- Procure AI
- Department
- Engineering
- Team
- Application Engineering
- Kind
- Decision-maker
- Approver
- Lead DBA
- Code
- ENG-07
It designs schema changes and migrations for the platform's transactional stores (requisitions, POs, receipts, invoices, contracts, supplier master), reviews slow queries such as match lookups and spend reports, and plans each step for lock time, table size and rollback. It decides whether a migration runs online now, waits for the maintenance window clear of client payment runs and month-end, or must be split.
How it works
7 steps, and a person signs off before step 7.
- 01Looks it up
Finds the record the request is about in the database's catalogue and query statistics: the table sizes, existing indexes, lock waits, replica lag and slowest query statistics for the database named, and the payment-run calendar..
- 02Reads the request
Works out what is being asked and what a complete answer needs.
- 03Checks the facts
Checks figures, dates and rules against the source records, not from memory.
- 04Forecasts, when it decides
Projects what each option will cost or return before it recommends one.
- 05Works it through
Prepares the draft, match, recommendation or reply, with its sources.
- 06Asks a person
Hands the work to Lead DBA with its reasoning before anything goes out.
- 07Acts
Once approved, it acts in the migration tool: it schedules the approved migration steps for the chosen window. and records what was done.
What it works with
Your knowledge, your systems, your people.
Knows
The database standards: naming and type conventions (exact decimals for money, tenant id on every table), the expand-and-contract playbook, the index policy, maintenance windows, the month-end and payment-run blackout calendar, and the retention rules for financial records.
Reads from
the database's catalogue and query statistics: the table sizes, existing indexes, lock waits, replica lag and slowest query statistics for the database named, and the payment-run calendar.
Acts in
the migration tool: it schedules the approved migration steps for the chosen window.
Hands back
An object with steps (a list, each with order, change, online true or false, expected_lock_seconds, reversible true or false and rollback), run_when (now, maintenance_window or split), index_proposals (a list, each with table, columns, reason and expected_gain_ms), pre_checks (a list) and data_risks (a list).
Asks a person when
The plan drops or rewrites financial data, holds a lock longer than 3 seconds, runs on a table above 20000000 rows, or runs in production outside Sunday 02:00 to 04:00 UTC, except the last three and first two working days of the month.
Settings
Your limits, set by your managers.
| Setting | Default | What it controls |
|---|---|---|
| Longest blocking lock allowed online (s) | 3 | A step expected to hold a blocking lock longer is not run online; a lock on invoices blocks matching. |
| Rows that make a table large | 20000000 | Tables above this get batched backfills and online index builds. |
| Slow query threshold (ms) | 500 | A query slower than this on average is reviewed. |
| Maintenance window | Sunday 02:00 to 04:00 UTC, except the last three and first two working days of the month | When steps that cannot run online are scheduled; month-end close is excluded. |
| Wait before dropping (days) | 30 | A column or table is dropped only this long after the code stops using it. |
| Retention for financial records (years) | 7 | PO, invoice, payment and audit-trail rows younger than this are never dropped or truncated. |
Limits are settings, not code: a manager changes them without a developer.
Rules it always follows
- Plan every change as expand and contract: add, backfill, switch the code, then remove; each step is reversible or says why not.
- A step expected to hold a blocking lock longer than 3 seconds is split or scheduled in Sunday 02:00 to 04:00 UTC, except the last three and first two working days of the month, and never during a client's payment run.
- On tables above 20000000 rows, build indexes online and backfill in batches with a pause between.
- Drop a column or table only 30 days after the code stops using it, after a verified backup, and never drop financial or audit-trail data younger than 7 years.
- Every new table carries the tenant id and every index used by a tenant query leads with it.
- Review queries slower than 500 ms and propose an index only when the gain outweighs its write cost on invoice and PO ingestion.
- Check the latest backup and replica lag before any production step and stop if either is out of policy.
See the Database Agent agent on your own work.
We’ll run it on a sample of your work and show you what your approver would see.