Uh oh!
There was an error while loading. Please reload this page.
feat(fetch): expose timings - #32647
Closed
Simon Knott (Skn0tt) wants to merge 5 commits into
Closed
Conversation
This comment has been minimized.
This comment has been minimized.
Simon Knott (Skn0tt)force-pushed
the
expose-apiresponse-timing
branch
from
September 17, 2024 13:56
3de1612 to
a7ea946Compare
This comment has been minimized.
This comment has been minimized.
Simon Knott (Skn0tt)force-pushed
the
expose-apiresponse-timing
branch
from
September 17, 2024 14:14
9c743d9 to
0eb69c3Compare
This comment has been minimized.
This comment has been minimized.
make responseEnd optional make it more similar to existing types missed one!
Simon Knott (Skn0tt)force-pushed
the
expose-apiresponse-timing
branch
from
September 17, 2024 14:24
0eb69c3 to
8de8383CompareSimon Knott (Skn0tt)
marked this pull request as ready for review
September 17, 2024 14:24
| - `domainLookupStart` <[float]> Time immediately before the browser starts the domain name lookup for the | ||
| resource. The value is given in milliseconds relative to `startTime`, -1 if not available. | ||
| - `domainLookupEnd` <[float]> Time immediately after the browser starts the domain name lookup for the resource. | ||
| - `domainLookupEnd` <[float]> Time immediately after the browser ends the domain name lookup for the resource. |
ContributorAuthor
There was a problem hiding this comment.
this is a drive-by fix - pretty sure it shouldn't say "start the domain name lookup"
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
| socket.on('secureConnect', () => { tlsHandshakeAt = monotonicTime(); }); | ||
| // socks / http proxy | ||
| socket.on('proxyConnect', () => { tcpConnectionAt = monotonicTime(); }); |
ContributorAuthor
There was a problem hiding this comment.
Adding timings to the protocol uncovered that tcpConnectionAt was undefined for requests over SOCKS and HTTPS Proxy. Turns out that the library we use for that doesn't emit the connect event, but the proxyConnect event instead.
This comment has been minimized.
This comment has been minimized.
| const endAt = monotonicTime(); | ||
| // spec: https://developer.mozilla.org/en-US/docs/Web/API/PerformanceResourceTiming | ||
| const timing: channels.ResourceTiming = { | ||
| startTime: startAt, |
Collaborator
There was a problem hiding this comment.
The docs say startTime is a wall time, not monotonic time.
| // spec: https://developer.mozilla.org/en-US/docs/Web/API/PerformanceResourceTiming | ||
| const timing: channels.ResourceTiming = { | ||
| startTime: startAt, | ||
| domainLookupStart: dnsLookupAt ? 0 : -1, |
Collaborator
There was a problem hiding this comment.
Why zero and not relativeTime()?
ContributorAuthor
There was a problem hiding this comment.
My understand is that the browser has some other steps like cache resolution before the DNS lookup, so there might be time between startTime and the DNS lookup start. On Node.js, I don't think there's anything between that - so it's zero, because we know the DNS lookup happens immediately after the request start.
| const timing: channels.ResourceTiming = { | ||
| startTime: startAt, | ||
| domainLookupStart: dnsLookupAt ? 0 : -1, | ||
| domainLookupEnd: dnsLookupAt ? dnsLookupAt! - startAt : -1, |
Collaborator
There was a problem hiding this comment.
relativeTime()?
| body | ||
| body, | ||
| timing, | ||
| responseEndTiming, |
Collaborator
There was a problem hiding this comment.
Why do we have two sets of timings now?
ContributorAuthor
There was a problem hiding this comment.
For browser requests, most of the timings are transmitted in the Response type, and then responseEndTiming arrives later in the requestFinished and requestFailed events. I opted to make this similar, so we have a big set of timings in timing and the final timing in responseEndTiming.
I thought about amending the ResourceTiming type instead, but then requestFinished would suddenly have the response end time both in responseEndTiming and in response.timing.responseEnd - that felt confusing.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
Contributor
Test results for "tests 1"2 failed 3 flaky35496 passed, 659 skipped Merge workflow run. |
Simon Knott (Skn0tt)
commented
Sep 27, 2024
ContributorAuthor
We discussed this with the team and decided against exposing this via the API. Playwright isn't a network performance testing tool, and we don't want people using it like one. Rough timings can easily be measured in userland. I'll open a separate PR to fix the bugs around HTTP / SOCKS Proxy we found in the existing measurements for HAR timings. |
Simon Knott (Skn0tt) added a commit
that referenced
this pull request
Oct 7, 2024
) Fixes a bug discovered in #32647. When using http proxy, the `connect` event isn't emitted so we don't populate `tcpConnectionAt`. The updated version of `https-proxy-agent` emits a `proxyConnect` as a replacement, so this PR updates and listens to that event. For socks proxies, the `on("socket")` event is emitted once the SOCKS connection is established, which is the equivalent of having a TCP connection available. --------- Signed-off-by: Simon Knott <info@simonknott.de> Co-authored-by: Max Schmitt <max@schmitt.mx>
Shahzad (shahzad31) added a commit
to elastic/synthetics
that referenced
this pull request
May 19, 2026
Playwright's APIResponse doesn't expose securityDetails(), serverAddr(), or request/response timings (upstream microsoft/playwright#32647 and microsoft/playwright#34938 were both declined; see microsoft/playwright#40905 for the current ask). To still surface this for HTTPS API monitoring use cases (cert expiry, remote-address alerting, response size tracking), open a side-channel tls.connect() in parallel with each request and fold the result into the NetworkInfo entry: - securityDetails: issuer, subjectName, protocol, validFrom, validTo - remoteIPAddress / remotePort - coarse dns / connect / ssl timings Probes are cached per host:port within a journey and silently skipped for HTTP or on failure (timeout, refused, untrusted) so they never affect the actual request. Also derive request and response body bytes (Content-Length preferred, buffer fallback) and emit server.ip / server.port in the JSON reporter. Co-authored-by: Cursor <cursoragent@cursor.com>
Shahzad (shahzad31) added a commit
to elastic/synthetics
that referenced
this pull request
Jun 15, 2026
Playwright's APIResponse doesn't expose securityDetails(), serverAddr(), or request/response timings (upstream microsoft/playwright#32647 and microsoft/playwright#34938 were both declined; see microsoft/playwright#40905 for the current ask). To still surface this for HTTPS API monitoring use cases (cert expiry, remote-address alerting, response size tracking), open a side-channel tls.connect() in parallel with each request and fold the result into the NetworkInfo entry: - securityDetails: issuer, subjectName, protocol, validFrom, validTo - remoteIPAddress / remotePort - coarse dns / connect / ssl timings Probes are cached per host:port within a journey and silently skipped for HTTP or on failure (timeout, refused, untrusted) so they never affect the actual request. Also derive request and response body bytes (Content-Length preferred, buffer fallback) and emit server.ip / server.port in the JSON reporter. Co-authored-by: Cursor <cursoragent@cursor.com>
Shahzad (shahzad31) added a commit
to elastic/synthetics
that referenced
this pull request
Jul 14, 2026
* fix(formatter): vendor JS codegen after Playwright 1.61 stopped exporting it Playwright 1.61 bundles its server code into coreBundle.js and no longer ships the codegen classes as importable modules, so the recorder's `playwright-core/lib/server/codegen/javascript` deep import no longer resolves. Vendor the minimal slice SyntheticsGenerator extends (_asLocator + _generateActionCall + JavaScriptFormatter) into src/formatter/codegen.ts, reusing the still-exported `iso` helpers (asLocator / formatObject / escapeWithQuotes) rather than vendoring the heavy locator logic. Formatter snapshots are unchanged. Also bump playwright/-chromium/-core to 1.61.0 (required for the native APIResponse TLS APIs) and refresh the device-descriptor Chrome UA in the options test that 1.61's bundled descriptors updated. Co-authored-by: Cursor <cursoragent@cursor.com> * feat: multi step api runner Introduces apiJourney() DSL plus APIDriver, APINetworkManager, and the type-aware Runner/Gatherer/PluginManager branching needed to run API-only journeys without launching Chromium. Co-authored-by: Cursor <cursoragent@cursor.com> * fix(api-journey): address review feedback, plug data loss, add tests Builds on the rebased api-journey commit: - PluginManager.output() now surfaces APINetworkManager results (previously dropped silently because of an instanceof NetworkManager check), and onStep() uses a proper union narrow instead of a lying cast. The browser/api branching is hidden behind a shared NetworkPlugin contract so the manager is transparent to journey type. - Runner only launches Chromium when at least one browser journey is scheduled; pure API suites skip launch entirely. - APIJourney now overrides _updateMonitor so 'synthetics push' registers it as an HTTP monitor instead of mislabeling it as browser. Journey base class gained a protected _setMonitor helper to make subclassing safe. - apiJourney.skip / apiJourney.only are now wired up. - APINetworkManager rewritten: only patches request.fetch (Playwright's helpers funnel through it, so the previous double-patch was double-counting requests), restores the prototype method on stop, handles fetch(Request, opts), wraps in try/finally so failed requests still leave a valid entry, drops dead Page/Frame barriers, surfaces status/headers/url/statusText. - APIJourney class trimmed: dead #cb / #driver fields removed; subclass now carries the http monitor override and forwards string|options upstream. - Reporter payload no longer carries browserDelay / browserconsole for API journeys; common_types updated accordingly. - Public API exports APIJourney / APIJourneyCallback / APIJourneyCallbackOpts / APIJourneyOptions / APIJourneyWithAnnotations. - Tests added: dsl/api-journey, plugins/api-network round-trip against a local HTTP server, plugins/plugin-manager API-driver coverage, core/api-runner end-to-end with browser launch spy, core/api-journey-register factory + skip/only wiring. Co-authored-by: Cursor <cursoragent@cursor.com> * feat(api-journey): capture TLS cert info, server.ip/port, and body bytes Playwright's APIResponse doesn't expose securityDetails(), serverAddr(), or request/response timings (upstream microsoft/playwright#32647 and microsoft/playwright#34938 were both declined; see microsoft/playwright#40905 for the current ask). To still surface this for HTTPS API monitoring use cases (cert expiry, remote-address alerting, response size tracking), open a side-channel tls.connect() in parallel with each request and fold the result into the NetworkInfo entry: - securityDetails: issuer, subjectName, protocol, validFrom, validTo - remoteIPAddress / remotePort - coarse dns / connect / ssl timings Probes are cached per host:port within a journey and silently skipped for HTTP or on failure (timeout, refused, untrusted) so they never affect the actual request. Also derive request and response body bytes (Content-Length preferred, buffer fallback) and emit server.ip / server.port in the JSON reporter. Co-authored-by: Cursor <cursoragent@cursor.com> * test(api-runner): cover empty steps, cookie isolation, and HTTPS e2e - Empty step with no requests must not break network-event step attribution: the next step's requests must still be assigned to that step's reference identity (required for waterfall grouping). - API journeys must have isolated APIRequestContexts: a cookie set in journey A must not leak into journey B. - HTTPS journey verifies the full pipeline emits TLS securityDetails, remote address, and response body bytes. Co-authored-by: Cursor <cursoragent@cursor.com> * docs(api-journey): add README sections and runnable example Splits the README usage section into "Browser journeys" and "API journeys (no browser)" so users discover apiJourney() without having to dig through the Elastic docs site. Adds a runnable example under examples/todos/api.journey.ts showing OAuth-style multi-step API checks. Co-authored-by: Cursor <cursoragent@cursor.com> * style(api-journey): fix lint and prettier issues - Replace non-null assertions with explicit narrowing in api-tls and api-runner tests so the `@typescript-eslint/no-non-null-assertion` rule is satisfied. - Apply prettier formatting to api-tls.ts and json.test.ts. Co-authored-by: Cursor <cursoragent@cursor.com> * fix(api-journey): support ES module inline scripts in the CLI loader Heartbeat hands API monitor scripts to `elastic-synthetics` as inline source via `--inline`. When that source uses ESM (e.g. `import` / top-level await) — which is the natural shape for `apiJourney()` and `step()` imports — Node's `vm.runInContext` path falls over with `SyntaxError: Cannot use import statement outside a module`, silently dropping the run. Detect ESM-shaped inline source and execute it via a temporary `.mjs` module file resolved with a `Module._resolveFilename` alias so `@elastic/synthetics` keeps resolving to the agent-installed copy. The CommonJS fast path is unchanged. Co-authored-by: Cursor <cursoragent@cursor.com> * test(api-journey): destructure params in inline ESM apiJourney test The new ESM inline loader path compiles the source as a regular module rather than running it through the new Function(...) wrapper, so `params` is no longer injected as an implicit local. The test needs to pull it out of the apiJourney callback args like the sibling browser test does. Co-authored-by: Cursor <cursoragent@cursor.com> * fix(api-journey): harden TLS reporter, IP-literal probe, inline loader Address review findings from end-to-end review of the API journey work: - src/reporters/json.ts: `formatTLS` previously called `new Date(undefined * 1000).toISOString()` when a TLS probe resolved with `protocol` set but cert dates missing (malformed `valid_from` / `valid_to`). That throws `RangeError: Invalid time value` and sinks the entire `journey/end` document. Route the dates through a defensive `epochToIso` helper that returns `undefined` for non-finite inputs, and add a regression test. - src/plugins/api-tls.ts: For IP-literal hosts the `lookup` event never fires, so `dnsEnd` stayed at its `-1` sentinel and cascaded into `connect: -1`. Treat the missing DNS phase as `dns: 0` and measure `connect` from `dnsStart`, so the timing breakdown stays meaningful. Add a probe test that exercises this path. - src/plugins/api-network.ts: `_currentStep` was typed `Partial<Step>` but initialised to `null`, contradicting the shared `NetworkPlugin` shape. Widen the field type to `Partial<Step> | null`. - src/loader.ts: Drop the `journey(` / `apiJourney(` heuristic from `isModuleInlineSource`. The regex matched inside string literals and comments, silently routing legacy inline scripts through the ESM loader and stripping the implicit `step` / `page` / `params` injection. Key off `import` / `export` only — that's the contract documented in the README. Also register a best-effort `process.exit` cleanup hook for the materialised `mkdtempSync` directory so long-lived hosts don't accumulate tmp dirs. - __tests__/plugins/api-network.test.ts: Add a guard that exercises every `APIRequestContext` helper (`get`/`post`/`put`/`patch`/ `delete`/`head`/`fetch`) so a future Playwright that bypasses `this.fetch` on any of them stops being a silent capture loss. - __tests__/core/api-runner.test.ts: Add a mixed-mode test verifying that a suite with both a `journey()` and an `apiJourney()` launches Chromium exactly once and routes each journey through the right driver type. Co-authored-by: Cursor <cursoragent@cursor.com> * docs(templates): add API journey scaffold examples Mirror the existing browser-journey scaffold examples with API-journey counterparts so users adopting `npx @elastic/synthetics <dir>` see both monitoring shapes side by side. - templates/journeys/api-example.journey.ts: minimal `apiJourney` with two GET steps and status assertions, structurally parallel to the existing `example.journey.ts`. - templates/journeys/advanced-api-example.journey.ts and its `advanced-api-example-helpers.ts`: multi-step API journey demonstrating the recommended shape — small reusable step builders, shared state populated by earlier steps and consumed by later ones, with a thunk-based id deletion to make the registration vs. execution timing explicit. - templates/synthetics.config.ts: adds `params.apiUrl` defaulting to jsonplaceholder.typicode.com so the API examples run out of the box; users override per-environment for their own service. - templates/README.md: distinguishes browser vs. API journey examples and explains when to reach for each. Co-authored-by: Cursor <cursoragent@cursor.com> * test(cli): unflake json-reporter test against chunked stdout `CLIMock.output()` returns only the last stdout chunk, and the existing test fed that into `JSON.parse`. On Linux CI a single chunk can hold multiple NDJSON events from the json reporter, which makes the parse throw `SyntaxError: Unexpected non-whitespace character after JSON at position N`. The test was therefore order- and flush-dependent and recently started flaking. Switch to the `cli.buffer()` accumulator (which join+split-by-line correctly reconstructs NDJSON regardless of chunk boundaries) and locate the `journey/start` event explicitly instead of assuming the last chunk holds exactly one event. A defensive `tryParse` swallows the rare partial-line case where the listener detaches mid-event, so the lookup keeps working without resorting to longer waits. Co-authored-by: Cursor <cursoragent@cursor.com> * refactor(api-journey): use native APIResponse TLS APIs from Playwright 1.61 Playwright 1.61 (microsoft/playwright#40932) exposes APIResponse.securityDetails() and APIResponse.serverAddr(), returning the same shapes the browser network path already consumes. Drop the tls.connect() side-channel (api-tls.ts), its per-origin cache, and the probe-fold blocks in APINetworkManager in favour of reading cert info and remote address straight off the response used by the actual request. This gives true per-request fidelity (final hop on redirects), removes the extra parallel TLS handshake, and now also reports server.ip/port over plain HTTP. The synthesized dns/connect/ssl timings (which came from a separate socket) are gone. A small normalizeTLSProtocol keeps the "TLSv1.3" -> "TLS 1.3" shape the JSON reporter expects. Co-authored-by: Cursor <cursoragent@cursor.com> * fix(api-journey): address review feedback and guard push by version - Register API journeys as `api` monitor type (was `http`) and bundle them like browser monitors in push (buildMonitorSchema + dry-run extraction). - Emit ECS `server.ip`/`server.port` for API journeys only; browser output keeps the address under `http.response`. - Use monotonic `now()` for API network timings. - Abort push with a clear message when API monitors target Kibana < 9.6.0. - Tighten verbose comments across the API journey code. Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: Cursor <cursoragent@cursor.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for freeto join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes#19621. Adds the same
timings()we have for browser responses toAPIResponse. Please apply some extra care in reviewing the timing calculations.In the protocol change, I was unsure wether to extend the existing
ResourceTimingtype or to add another property. I went with the added property to be in-line with how the events work for browser requests - let me know if we should do it differently instead.