* Centralize the retry package for the otlp exporters
* Add dependabot config
* Use unified internal/retry module in otlp exporters
* Remove otlpmetric and otlptrace internal/retry module
* Run go mod tidy in otlp{metric,trace}
* Use gRPC ClientConn for otlpmetricgrpc conn handling
* Update PR number
* Update otlpmetrichttp to use retry package
* Remove old commented code
* Add all other external changes to changelog
* POC using the grpc.ClientConn to handle connections
* Update invalid client security test
* Update client start test for a bad endpoint
* Use any ClientConn a user provides
* Connect ReconnectionPeriod to gRPC conn retries
* Replace connection retry handling direct in otlptracegrpc
* Fix client comments
* Fix comment for NewGRPCConfig
* Replace reconnection test
* Fix grammar
* Remove unrelated changes
* Remove connection pkg
* Rename evaluate to retryable
* POC using the grpc.ClientConn to handle connections
* Replace connection retry handling direct in otlptracegrpc
* Add ClientConn use changes to changelog
* Update otlptracegrpc options
* Only close ClientConn that the Client create
* Remove listener wrapper from mock_collector_test
This is not needed now that no tests relies on the listener to wait for
a connection to be established before continuing.
* Fix spelling error
* Do not use deprecated options in the otel-collector example
* Add unit tests for retryable and throttleDelay funcs
* Add unit tests for context heredity
* Add test that exporter stop is linked to context cancel
* go mod tidy
* Update exporters/otlp/otlptrace/otlptracegrpc/client.go
Co-authored-by: Anthony Mirabella <a9@aneurysm9.com>
* Fix go.mod from rebase
* Remove wrong comment about client stop closing gRPC conn
* Fix shutdown test cleanup
Do not check the second call to the client Stop. There is no guarantee
it will not error in normal operation.
* Make lint fixes
* Fix flaky unit test
Use the internals of the client to explicit cancel the context returned
from exportContext. This gets around the bug where the select in Stop
may randomly choose the non-context Done case and avoid returning an
error (also failing to cancel the context).
* Remove deprecation
To configure the client/exporter with environment variables these
options are used. There is no way to fully remove these options without
removing support for configuration with environment variables. Leave
that decision and strategy determination to a separate PR.
* Fix grammatical error in comment
Co-authored-by: Anthony Mirabella <a9@aneurysm9.com>
* Add a go version compatibility to the readme.
* Fix lint
* Update README.md
Co-authored-by: Robert Pająk <rpajak@splunk.com>
Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com>
Co-authored-by: Robert Pająk <rpajak@splunk.com>
* Added Benchmarks around events
Also fixed the attribute benchmarks to reflect the # of attributes in the test
* Fix test txt, and too new a version of time
Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com>
* Fail in RunEndToEndTest if collector stop fails
* Use testing T.Cleanup to check shut downs
* Add goroutine leak detection
* Fix TestExporterShutdown go leak
The shutdown tests checking if a context error is honored did not
completely clean up the resources used by the client after the error was
evaluated. Update the connection client to handle multiple calls to
shutdown and make a second call to these clients that must succeed so
the test does not have abandoned goroutines.
* Fix leak in TestNew_WithTimeout
The mockTraceService did not delay with its lock being held. This
resulted in the mockCollector stopping and being able to acquire the
lock. It was assumed that no export was taking place because of this and
the mockTraceService was abandoned without cleaning up resources it held
and goroutines it had spawned. This reworks the export blocking logic to
block on a channel read. This will make the block more deterministic and
not depend on the scheduler timing. Additionally, this blocking is moved
inside the lock acquire. Meaning code will deadlock if the block is not
released before a shutdown (something the developer will immediately be
aware of when they submit a bad patch), and will ensure all resources
are released before shutdown.
Replace TestNew_WithTimeout with TestExportSpansTimeoutHonored which
directly tests if a span export errors when the timeout is reached. This
is the only unique thing that TestNew_WithTimeout, but it also tests the
non-error path. That non-error path is tested in many other tests.
* Guard otlptracehttp client stopCh when stopping
In normal operations the exporter is guaranteed to only ever call the
client Stop method once. However in testing we need to call this
multiple times when checking it returns an error in particular context.
Add a lightweight sync.Once to the closing of the stopCh to ensure tests
do not panic when cleaning up.
* Release export block after export
Prevent deadlock in TestExportSpansTimeoutHonored.
* [website_docs] Link to website page via path
Contributes to https://github.com/open-telemetry/opentelemetry.io/issues/922. /cc @austinlparker
* Make links relative and wrap paragraph text
Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com>
Remove nil check on return from NewTestSpanProcessor as it can never be
nil, addressing #2396. Also, add nil checks for testSpanProcessor
methods to prevent panics.
* change resource.Default to be evaluated on first call, rather than on import
* guard defaultResource creation with sync.Once, fallback to emtpyResource
* update CHANGELOG
Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com>
This is a simple loggin interface to be used interally for the sdk.
Included is an exmaple in `namedtracer` run:
```
cd example/namedtracer
go run . >/dev/null
```
Bump github.com/golangci/golangci-lint from 1.42.1 to 1.43.0 in /internal/tools.
Bump google.golang.org/grpc from 1.41.0 to 1.42.0 in /exporters/otlp/otlpmetric.
Bump go.opentelemetry.io/proto/otlp from 0.9.0 to 0.10.0 in /exporters/otlp/otlpmetric/otlpmetrichttp.
Bump go.opentelemetry.io/proto/otlp from 0.9.0 to 0.10.0 in /exporters/otlp/otlpmetric.
Bump go.opentelemetry.io/proto/otlp from 0.9.0 to 0.10.0 in /exporters/otlp/otlptrace.
Bump google.golang.org/grpc from 1.41.0 to 1.42.0 in /exporters/otlp/otlpmetric/otlpmetricgrpc.
Bump google.golang.org/grpc from 1.41.0 to 1.42.0 in /exporters/otlp/otlptrace.
Bump go.opentelemetry.io/proto/otlp from 0.9.0 to 0.10.0 in /exporters/otlp/otlptrace/otlptracehttp.
Bump google.golang.org/grpc from 1.41.0 to 1.42.0 in /exporters/otlp/otlptrace/otlptracegrpc.
Bump go.opentelemetry.io/proto/otlp from 0.9.0 to 0.10.0 in /exporters/otlp/otlpmetric/otlpmetricgrpc.
Bump go.opentelemetry.io/proto/otlp from 0.9.0 to 0.10.0 in /exporters/otlp/otlptrace/otlptracegrpc.
Bump github.com/benbjohnson/clock from 1.1.0 to 1.2.0 in /sdk/metric.
Bump google.golang.org/grpc from 1.41.0 to 1.42.0 in /example/otel-collector.
* Empty queued spans when ForceFlush called
Update the implementation of ForceFlush() to first ensure that all spans
which are queued are added to the batch before calling export spans.
Create a small ReadOnlySpan implementation which can be used as a marker
that ForceFlush has been invoked and used to notify when all spans are
ready to be exported.
Fixes#2080.
* Add a changelog entry.
* Update CHANGELOG.md
Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com>
* Update sdk/trace/batch_span_processor.go
Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com>
* Improve test case to verify multiple flushes.
* Refactor code to use enqueue.
* Be more defensive on waiting for queue.
Update the handling of the force flush span so we only wait on the
channel if we were able to enqueue the span to the queue.
* Fix linter.
Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com>
Co-authored-by: Anthony Mirabella <a9@aneurysm9.com>
