[EXPORTER] Add custom HTTP client by ltowarek · Pull Request #4071 · open-telemetry/opentelemetry-cpp

ltowarek · 2026-05-08T21:21:24Z

Fixes #2084

Changes

Adds a way to provide a custom HTTP client instead of the default libcurl-based client.

The main motivation are embedded and cross-compiled targets where libcurl is unavailable. A working example for ESP32S3 using ESP-IDF's esp_http_client as the transport backend is at ltowarek/esp-opentelemetry-cpp#30.

linux-foundation-easycla · 2026-05-08T21:21:31Z

The committers listed above are authorized under a signed CLA.

✅ login: ltowarek / name: Lukasz Towarek (a323792408764f461e2d16828358376c59f3dacb)
✅ login: ThomsonTan / name: Tom Tan (07d64830d450772c6a829e7ed921a5a804d05c1e, 8b2112a72538927dbcf8d6ae73d450fdc399df15, a5264df01d0ab7e5abfaa4c229a3c378c15a0a91)

owent · 2026-05-11T13:46:31Z

Could you please add a test to use a simple custom http client?

lalitb · 2026-05-11T17:29:29Z

Thanks @ltowarek. Couple of review comments as of now -

As @owent mentioned, could we also add a minimal example for this option, even if the transport is just a no-op custom HttpClient / HttpClientSync? The important part is showing where the user provides HttpClientFactory::Create() and CreateSync(), and how their target is linked with the OTLP HTTP exporter.
Also, I believe there are still a few CMake paths that bring in curl directly. For example, Zipkin, Elasticsearch, HTTP examples, and the W3C tracecontext HTTP test server still link opentelemetry_http_client_curl / CURL::libcurl, while WITH_CUSTOM_HTTP_CLIENT=ON disables the curl client target globally. Can we either make this option explicitly incompatible with those curl-only paths, or update them to follow the same custom-client path?

codecov · 2026-05-11T17:49:01Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 82.56%. Comparing base (c98d949) to head (aba37bf).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #4071      +/-   ##
==========================================
+ Coverage   82.08%   82.56%   +0.49%     
==========================================
  Files         386      389       +3     
  Lines       16211    16617     +406     
==========================================
+ Hits        13305    13718     +413     
+ Misses       2906     2899       -7

Files with missing lines	Coverage Δ
...y/exporters/elasticsearch/es_log_record_exporter.h	`100.00% <ø> (+100.00%)`	⬆️
...orters/elasticsearch/src/es_log_record_exporter.cc	`12.22% <100.00%> (+12.22%)`	⬆️
...de/opentelemetry/exporters/otlp/otlp_http_client.h	`100.00% <ø> (ø)`
exporters/otlp/src/otlp_http_client.cc	`67.50% <100.00%> (-0.06%)`	⬇️
exporters/otlp/src/otlp_http_exporter.cc	`99.56% <100.00%> (+0.51%)`	⬆️
exporters/otlp/src/otlp_http_exporter_factory.cc	`78.95% <100.00%> (+15.32%)`	⬆️
...xporters/otlp/src/otlp_http_log_record_exporter.cc	`99.12% <100.00%> (+0.99%)`	⬆️
.../otlp/src/otlp_http_log_record_exporter_factory.cc	`78.95% <100.00%> (+15.32%)`	⬆️
exporters/otlp/src/otlp_http_metric_exporter.cc	`99.14% <100.00%> (+0.96%)`	⬆️
...ters/otlp/src/otlp_http_metric_exporter_factory.cc	`78.95% <100.00%> (+15.32%)`	⬆️
... and 5 more

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

dbarker

Thanks for the PR. Please see the question below.

ltowarek · 2026-05-12T21:02:22Z

Thanks for the feedback! I think I've covered everything in the new patch set:

Added opentelemetry_http_client interface library to easily inject different implementations.
Added tests based on http_client::nosend::HttpClient.
Added an example app.
Used WITH_HTTP_CLIENT_CURL=OFF instead of WITH_CUSTOM_HTTP_CLIENT.

marcalff

Thanks for the contribution.

Please consider to use:

an abstract class HttpClientFactory, with virtual methods for Create and CreateSync
A child class HttpCurlClientFactory, which implement this for CURL
a HttpCurlClientFactory::singleton

Everywhere the static method HttpClientFactory::Create() is used today, pass a HttpClientFactory instance instead, taken as input parameter.

The user of an HTTP client can then decide which implementation to use:

pass HttpCurlClientFactory::singleton to use CURL
pass an instance of MyOwnHttpClientFactory for alternate implementations.

It should be possible to mix CURL and non CURL HTTP clients in the same binary, for example:

send traces using HTTP + CURL to endpoint A
send metrics using HTTP + alternate to endpoint B

ltowarek · 2026-05-13T21:51:02Z

Thanks for the contribution.

Please consider to use:
* an abstract class `HttpClientFactory`, with virtual methods for `Create` and `CreateSync`

* A child class `HttpCurlClientFactory`, which implement this for CURL

* a HttpCurlClientFactory::singleton
Everywhere the static method HttpClientFactory::Create() is used today, pass a HttpClientFactory instance instead, taken as input parameter.

The user of an HTTP client can then decide which implementation to use:
* pass HttpCurlClientFactory::singleton to use CURL

* pass an instance of MyOwnHttpClientFactory for alternate implementations.
It should be possible to mix CURL and non CURL HTTP clients in the same binary, for example:
* send traces using HTTP + CURL to endpoint A

* send metrics using HTTP + alternate to endpoint B

@marcalff, it looks like a good idea.

My main goal with the current PR is build-time configuration - to enable opentelemetry-cpp for embedded systems as they simply can't build
libcurl. The binary for such systems will always contain only one HTTP client.

I think runtime changes should be a part of a separate PR which I'm open to working on.

ltowarek · 2026-05-14T21:45:13Z

@marcalff I've added runtime changes to this PR. Please let me know if b8eb45d fits the needs.

lalitb · 2026-05-15T01:06:54Z

Thanks for adding the runtime factory path. I think this is moving in the right direction, but can we make the injection story consistent across all HTTP exporters?

Right now OtlpHttpExporterFactory has overloads that accept a custom HttpClientFactory / HttpClient, but the OTLP HTTP log and metric exporter factories still only expose the default-client path. That means a user can inject a custom client for traces, but not for logs/metrics through the same public factory API.

Also, after making HttpClientFactory::Create*() virtual instance methods, there still seem to be a few static call sites left, for example the HTTP example and test_ext_http_curl.cc. Those should probably go through GetDefaultHttpClientFactory()->Create*() or a concrete curl factory.

…/opentelemetry-cpp into fix/issue-24-custom-http-client

marcalff

Thanks for the contribution.

This is good work.

I suspect GetDefaultHttpClientFactory() may need minor adjustments later, but this will be for a different PR.

dbarker

Thanks for the feature! Please see feedback below to build/test the new example in CI. Once that is in and CI is passing this should be good to merge.

Replaces the libcurl cross-compile workaround (stub FindCURL.cmake, ~30 CURL_*/SIZEOF_*/HAVE_* cache variables, and the opentelemetry_http_client_curl target hijack) with WITH_HTTP_CLIENT_CURL=OFF and the new OtlpHttpExporter(options, HttpClient) constructor, passing MakeEspHttpClient() directly. Bumps third_party/opentelemetry-cpp to 9e93606d, which includes the merged open-telemetry/opentelemetry-cpp#4071.

* Remove CURL workarounds * Use WITH_HTTP_CLIENT_CURL=OFF; bump opentelemetry-cpp Replace the non-existent WITH_CUSTOM_HTTP_CLIENT flag with the correct WITH_HTTP_CLIENT_CURL=OFF, which opentelemetry-cpp uses to signal that the caller supplies its own HttpClientFactory transport. Bump third_party/opentelemetry-cpp to e568cc0 (tip of the local fix/issue-24-custom-http-client fork branch), which adds the opentelemetry_http_client interface library and the custom_http_client example that accompanies opentelemetry-cpp PR #4071. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * Refactor HttpClientFactory to use EspHttpClientFactory for improved clarity and maintainability * Remove stale libcurl references from comments Update comments in the transport source, header, and README to reflect that libcurl is now excluded entirely via WITH_HTTP_CLIENT_CURL=OFF rather than replaced at the link level. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * Use opentelemetry-cpp#4071 custom HTTP client instead of curl Replaces the libcurl cross-compile workaround (stub FindCURL.cmake, ~30 CURL_*/SIZEOF_*/HAVE_* cache variables, and the opentelemetry_http_client_curl target hijack) with WITH_HTTP_CLIENT_CURL=OFF and the new OtlpHttpExporter(options, HttpClient) constructor, passing MakeEspHttpClient() directly. Bumps third_party/opentelemetry-cpp to 9e93606d, which includes the merged open-telemetry/opentelemetry-cpp#4071. --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>

…ccepts Main grew its own FakeResponse, FakeRequest, FakeSession and FakeHttpClient in open-telemetry#4071 and open-telemetry#4501, in an unnamed namespace, and this branch already had four of the same names in another unnamed namespace in the same file. Reopening an unnamed namespace names the same namespace, so the rebase merged both without a conflict into four redefinitions that do not compile. Merged rather than renamed. Main's set is used once and answers with one fixed response; this branch's runs a script the case supplies, which is the general case. So the session and the client take a script and default to answering the way main's did, and the existing call site is untouched. The default body had to change too, which is not cosmetic. It read `{"errors": false, "failed" : 0}`, and this pull request stops reading `failed` and starts requiring one `items` result per submitted record, so that body is now rejected: with it, ExportingARecordWithInvalidUtf8DoesNotAbort returns kFailure and fails. The default is now a body a server could send for a one record batch, and the comment that described the old substring search is rewritten. Verified in both configurations, since the guards cut different code: 26 tests, 22 passed and 4 skipped without async export, 20 passed and 6 skipped with it, no failures in either. Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>

Main's open-telemetry#4071 and open-telemetry#4501 put FakeResponse, FakeRequest, FakeSession and FakeHttpClient in an unnamed namespace at the top of this file, and this branch already had four of those names in a second unnamed namespace lower down. Reopening an unnamed namespace names the same namespace, so the rebase merged both with no conflict at all and left four redefinitions. One set now. The session and the client take a script and default to answering the way main's did, so main's own call site needs no edit, and DeferredSession and DeferredHttpClient are untouched. The default body stays as main wrote it. This branch does not change how the exporter decides success, so what main's cases send still passes here. Verified in both configurations with maintainer mode on: 19 cases, all passing without async export, 3 passing and 16 skipping with it. Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>

Main's open-telemetry#4071 and open-telemetry#4501 put FakeResponse, FakeRequest, FakeSession and FakeHttpClient in an unnamed namespace at the top of this file, and this branch already had four of those names in a second unnamed namespace lower down. Reopening an unnamed namespace names the same namespace, so the rebase merged both with no conflict at all and left four redefinitions. One set now. The session and the client take a script and default to answering the way main's did, so main's own call site needs no edit. The script carries the handler as a shared_ptr rather than a reference, because the cases here have to keep it and send a second event to it, and the client keeps its on_create_session and on_cancel_all hooks, both empty by default. The default body stays as main wrote it. This branch does not change how the exporter decides success, so what main's cases send still passes here. Verified in both configurations with maintainer mode on: 12 cases, 12 passing with async export and 3 passing with 9 skipping without it. Removing CompleteOnce's compare and exchange turns six of them red, so the rewritten fixtures still discriminate. Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>

ltowarek marked this pull request as ready for review May 8, 2026 21:40

ltowarek requested a review from a team as a code owner May 8, 2026 21:40

dbarker reviewed May 12, 2026

View reviewed changes

Comment thread exporters/otlp/CMakeLists.txt Outdated

marcalff previously requested changes May 12, 2026

View reviewed changes

ltowarek added 8 commits May 13, 2026 23:37

[EXPORTER] Add WITH_CUSTOM_HTTP_CLIENT option

72f8cce

Fix formatting

8fa2072

Add opentelemetry_http_client interface library

87a12da

Add example and test

135c213

Use WITH_HTTP_CLIENT_CURL=OFF

53debbe

Improve the custom_http_client example

33f0e4d

Link opentelemetry_exporter_otlp_http_client with CURL

fba51be

Include custom_http_client example only in build-time

516c51b

ltowarek force-pushed the fix/issue-24-custom-http-client branch from 5147826 to 516c51b Compare May 13, 2026 21:38

owent reviewed May 14, 2026

View reviewed changes

Comment thread exporters/zipkin/CMakeLists.txt Outdated

Install http_client

235c596

ltowarek changed the title ~~[EXPORTER] Add WITH_CUSTOM_HTTP_CLIENT option~~ [EXPORTER] Add custom HTTP client May 14, 2026

ltowarek requested review from marcalff and owent May 14, 2026 18:17

Make HttpClientFactory virtual

b8eb45d

ltowarek added 3 commits May 15, 2026 21:52

Fix http_client_factory_curl.h

b424e86

Support HttpClientFactory in logs and metrics

5695f7d

Remove old HttpClientFactory calls

db50030

ltowarek requested a review from lalitb May 26, 2026 20:54

dbarker and others added 9 commits May 26, 2026 21:26

Merge branch 'main' into fix/issue-24-custom-http-client

da2f405

Merge branch 'main' into fix/issue-24-custom-http-client

e467bfa

Merge branch 'main' into fix/issue-24-custom-http-client

610bfe1

Make GetDefaultHttpClientFactory private

1bd6714

Fix ifdef

ae521e5

Merge branch 'main' into fix/issue-24-custom-http-client

19d9b57

Fix CI

9016ef2

Merge branch 'fix/issue-24-custom-http-client' of github.com:ltowarek…

21f1133

…/opentelemetry-cpp into fix/issue-24-custom-http-client

Merge branch 'main' into fix/issue-24-custom-http-client

6d5a8aa

marcalff approved these changes Jun 5, 2026

View reviewed changes

marcalff added 2 commits June 7, 2026 18:43

Merge branch 'main' into fix/issue-24-custom-http-client

bb1c8ce

Merge branch 'main' into fix/issue-24-custom-http-client

3447790

marcalff self-assigned this Jun 9, 2026

marcalff added 2 commits June 9, 2026 21:53

Merge branch 'main' into fix/issue-24-custom-http-client

5e8bbcb

Merge branch 'main' into fix/issue-24-custom-http-client

6c3c771

dbarker reviewed Jun 11, 2026

View reviewed changes

Comment thread examples/custom_http_client/CMakeLists.txt

Comment thread examples/CMakeLists.txt Outdated

Add example to the CI

e1b20e3

dbarker reviewed Jun 11, 2026

View reviewed changes

Comment thread examples/custom_http_client/CMakeLists.txt Outdated

Fix CI

5fa52bc

dbarker approved these changes Jun 11, 2026

View reviewed changes

Merge branch 'main' into fix/issue-24-custom-http-client

aba37bf

marcalff merged commit 9e93606 into open-telemetry:main Jun 12, 2026
71 checks passed

claude Bot mentioned this pull request Jun 14, 2026

Bump third_party/opentelemetry-cpp from 9e93606 to 767f797 ltowarek/esp-opentelemetry-cpp#44

Closed

BrewTestBot mentioned this pull request Jul 16, 2026

opentelemetry-cpp 1.28.0 Homebrew/homebrew-core#293633

Merged

1 task

Uh oh!

Conversation

ltowarek commented May 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Changes

Uh oh!

linux-foundation-easycla Bot commented May 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

owent commented May 11, 2026

Uh oh!

lalitb commented May 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

codecov Bot commented May 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

dbarker left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

ltowarek commented May 12, 2026

Uh oh!

marcalff left a comment

Choose a reason for hiding this comment

Uh oh!

ltowarek commented May 13, 2026

Uh oh!

Uh oh!

ltowarek commented May 14, 2026

Uh oh!

lalitb commented May 15, 2026

Uh oh!

marcalff left a comment

Choose a reason for hiding this comment

Uh oh!

dbarker left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

ltowarek commented May 8, 2026 •

edited

Loading

linux-foundation-easycla Bot commented May 8, 2026 •

edited

Loading

lalitb commented May 11, 2026 •

edited

Loading

codecov Bot commented May 11, 2026 •

edited

Loading