Watch live SQL
The Live SQL lens in proxymock web streams your app's database calls as they are recorded, the way SQL Server Profiler streams a trace. Run a request in your app and its statements appear a moment later, with who ran them, how long they took and how they ended. It works on a local proxymock record in progress and on a live tap of a Kubernetes workload, and it needs proxymock 2.5.1149 or later.
Use it when you want to see what a change does to the database while you click through the app: which statements a page runs, whether a loop fires one query per row, which call is slow, or which one fails.
Open the lens
- Start recording your app with its database mapped, as described in Record your database.
- Run
proxymock weband open Requests. - Pick the run that is recording from the Run list and choose Live SQL next to Grid, Report and Trace.
The lens shows the most recent 5,000 calls. The line at the top right counts every call that arrived since you opened it (seen), the database calls among them (database), the ones that pass your filters (shown), and how many older rows were dropped to stay within the buffer.
Columns
| Column | What it shows |
|---|---|
| Start | When the call started |
| Event | Statement, Login, Logout or Error, or why the call needs attention, such as a deadlock |
| Statement | The statement with values masked, after the route the app tagged it with when it did |
| App, User, DB, Session | Who ran it: the client application name, the database user, the database, and the server's session id |
| Duration, Rows | How long it took and how many rows it returned or changed |
| Status | The command tag, or the error code |
Protocol chatter such as SSL negotiation and describe-only messages counts as database traffic in the line at the top but is never listed.
Templates
A template narrows the lens to one kind of call, like Profiler's trace templates:
| Template | Shows |
|---|---|
| Standard | Every database call |
| Slow statements | Duration of 100 ms or more |
| Errors | Calls that returned a database error |
| Cancels and disconnects | Cancelled, timed out and terminated calls, cancel requests and dropped connections |
| Logins and logouts | Login and logout events |
| Locks | Deadlocks and lock timeouts |
| Big results | More than 1,000 rows |
A template sets the Duration and Database rows of the Requests filters and keeps the rest of your filters, so you can combine Slow statements with a service or a table. Edit the filters afterwards to make it your own; the template menu then reads Custom.
Follow, pause and clear
- Follow keeps the newest row in sight. Scrolling up turns it off, and scrolling back to the bottom turns it on again.
- Pause freezes the view while recording continues; the state shows how many calls are waiting. Resume shows them.
- Clear empties the view and keeps listening, so you can start a fresh look before you click the next button in your app.
Look at one call
Select a row to see a strip with its event, session, route, user and database, and links to act on it:
- Open request opens the call in the Requests grid. Connection opens it on the Connection tab, with every call on the same database connection. See Inspect a statement.
- Copy as SQL copies the statement as runnable SQL, with its bound values filled in. It is offered on the row that ran the statement; in the extended protocol a statement also arrives as Parse and Bind rows, which have nothing to run on their own.
- Filter: session, Filter: this statement and Filter: this route narrow the lens to calls like this one.
Keep what you caught
- Save as recording writes the calls in the lens to a new recorded run named
recorded-live-sql-<time>, which you can mock, replay or compare like any other recording. Protocol chatter is left out, so the run holds statements, logins and logouts. - Copy as SQL in the toolbar copies the rows on screen, in recorded order, as one script that
psqlormysqlcan run, with bound values filled in. Statements whose values were not captured are written commented out. When the clipboard is not available the script downloads instead. It needs proxymock 2.5.1154 or later. See Export a SQL script for the whole run.
Watch a cluster workload
For a workload in Kubernetes, start a live tap on it from Observability in proxymock web. The tap writes the workload's traffic into a run on your machine as it arrives, and the Live SQL lens follows that run the same way it follows a local recording.