Profiling HTTP Requests with the CLI¶
Blackfire lets you profile HTTP requests like web pages, web service calls, or API calls.
To get started, check that you have Blackfire installed correctly on the server hosting the website you want to profile.
Profiling an HTTP request means instrumenting the project’s code. Blackfire is apart from many other profilers because of two main differences:
- Blackfire automatically instruments your code when needed, so you don’t need to change your code;
- Blackfire does nothing when you are not profiling a request (the overhead is then almost zero), so it is safe to deploy Blackfire on production servers, even for heavy-traffic projects.
This cookbook describes how to profile from the command line interface. You can also profile web applications from a browser.
Profiling Simple HTTP Requests¶
Profiling an HTTP request can be done on the command line thanks to the
blackfire utility.
The easiest way to profile an HTTP request is to use the curl sub-command
of the blackfire utility. It accepts the same arguments and options as the regular curl utility:
1 | blackfire curl http://example.com/ |
Warning
Recommended: To explicitly define which
environment the profile must be sent
to, use the --env=<ENV_UUID> parameter.
v2.14.0+ of the unified Agent and Client:
when no target environment is specified, the Agent uses the
BLACKFIRE_SERVER_ID environment variable to determine where to send the
profile.
When no target environment is specified or implicitly discovered, the profile
is sent to the personal sandbox environment of the user whose
personal credentials (BLACKFIRE_CLIENT_ID
and BLACKFIRE_CLIENT_TOKEN) were used. Only that user can then access
the profile from /my/profiles.
To explicitly specify your personal sandbox environment, use
--env="My Environment"
You can get a list of options available for the curl sub-command:
1 | blackfire help curl |
To get more accurate results, take several samples of the request via the
--samples option (we recommend you to use this option only for “safe” HTTP
requests like GET requests):
1 | blackfire --samples 10 curl http://example.com/ |
At the end of the request, blackfire outputs the URL where the profile can
be found (hide it by passing the -q option.)
JSON Output¶
You can integrate Blackfire results into your tools by using the --json
option to get a JSON representation of a profile:
1 | blackfire --json curl http://example.com/ |
Note
The JSON output is displayed on STDOUT, while the regular output (e.g.
progress, profile summary…) is displayed on STDERR.
The resources consumed are available under the envelope entry;
keys are the cost dimensions:
- pmu: Peak Memory Usage (in bytes);
- nw: Network (in bytes);
- wt: Wall Clock Time (in microseconds);
- cpu: CPU time (in microseconds);
- io: I/O time (in microseconds).
Profiling non-GET Requests, Ajax, and More¶
Profiling non-GET requests or requests which need some specific HTTP headers is
no different as blackfire curl supports all cURL options:
1 | blackfire curl -XPOST http://example.com/ |
Profiling Part of an HTTP Call¶
Blackfire automatically instruments your code, but sometimes, you might want to focus the profiling on only part of the code. That’s possible when opting for manual instrumentation via the PHP SDK.
After instrumenting your code, use the blackfire utility as above to
profile your application. When not using Blackfire, all calls are converted to
no-ops.
Profiling all Requests While you Browse Requires Development/Production¶
Profiling POST requests, Ajax requests, and API calls can be done via the “Profile all Requests” link in the browser extension.
When profiling from a browser, select “Profile all requests”, then “Record!”. Blackfire will instrument all requests until you click on the “Stop” button in the pop-over.
A pop-over window will open at the bottom right of your screen, where all profiled requests will be listed.
All PHP requests which have been profiled will be shown on your dashboard.