census-instrumentation / opencensus-cpp Goto Github PK

View Code? Open in Web Editor NEW

145.0 25.0 74.0 989 KB

A stats collection and distributed tracing framework

Home Page: https://opencensus.io/

License: Apache License 2.0

Python 0.40% C++ 83.07% Shell 1.15% C 0.28% Batchfile 0.34% CMake 7.80% Dockerfile 0.02% Starlark 6.94%

stats opencensus trace

opencensus-cpp's People

Contributors

Stargazers

Watchers

opencensus-cpp's Issues

Add a Prometheus exporter

We want to integrate support for exporting stats to Prometheus as well as Stackdriver. Prometheus uses a pull model--clients expose stats on an HTTP endpoint that the server scrapes--which will require an extension to the StatsExporter API. For exposing stats, I see two sensible options:

Integrate with the existing Prometheus client library (https://github.com/jupp0r/prometheus-cpp).
Expose a function returning formatted exports as a string, leaving it to application developers to expose it via HTTP.

The second minimizes our dependencies, but duplicates formatting code from the existing client library and makes it difficult for applications using the Prometheus client directly to integrate the two--they would have to implement their own Exposer combining the two sources of metrics. I prefer the first--if we use the existing client library and make the PrometheusExporter a custom Collectable, users can use the default Exposer for out-of-the-box export (supporting combining Opencensus and native Prometheus metrics), while still having the option of writing their own handler (using the client's formatters) to avoid running the Prometheus client's default server.

Use absl::Span instead of initializer_list

I think it is better to use absl::Span for the AttributesRef definition because can be constructed from a initializer_list as well as from a vector or array.

This may be useful for cases where the attributes are determined at runtime.

Use std::move reduce BucketBoundaries copy overhead

I found BucketBoundaries when construction is through internal will make a copy, here to move construction will be better

BucketBoundaries(absl::Span<const double> lower_boundaries)
      : lower_boundaries_(lower_boundaries.begin(), lower_boundaries.end()) {}

  // The lower bound of each bucket, excluding the underflow bucket but
  // including the overflow bucket.
  std::vector<double> lower_boundaries_;

Call: 
  std::vector<double> boundaries(num_finite_buckets + 1);
  for (int i = 0; i <= num_finite_buckets; ++i) {
   ..........
  }
  return BucketBoundaries(boundaries);

replace

  BucketBoundaries(std::vector<double>&& lower_boundaries)
  std::vector<double> boundaries(num_finite_buckets + 1);
  for (int i = 0; i <= num_finite_buckets; ++i) {
   ..........
  }
  return BucketBoundaries(std::move(boundaries));

Build failure with -Wdefaulted-function-deleted

Because Context has a default move assignment operator but its data member Span explicitly deletes it.

Example build failure:
INFO: From Compiling external/io_opencensus_cpp/opencensus/stats/internal/recording.cc:
Step #4: In file included from external/io_opencensus_cpp/opencensus/stats/internal/recording.cc:21:
Step #4: In file included from external/io_opencensus_cpp/opencensus/tags/context_util.h:18:
Step #4: external/io_opencensus_cpp/opencensus/context/context.h:52:12: warning: explicitly defaulted move assignment operator is implicitly deleted [-Wdefaulted-function-deleted]
Step #4: Context& operator=(Context&&) = default;
Step #4: ^
Step #4: external/io_opencensus_cpp/opencensus/context/context.h:76:27: note: move assignment operator of 'Context' is implicitly deleted because field 'span_' has a deleted move assignment operator
Step #4: opencensus::trace::Span span_;
Step #4: ^
Step #4: external/io_opencensus_cpp/opencensus/trace/span.h:121:9: note: 'operator=' has been explicitly marked deleted here
Step #4: Span& operator=(Span&&) = delete;
Step #4: ^
Step #4: 1 warning generated.

Avoid copying protos in this repo from googleapis

Currently the google/ directly copies all the protos from googleapis repo. We should not do that and probably build from head.

trace/Span: SpanKind is missing

I am working on another iteration on the website that shows users each concept and usage in the various languages and so far for Spans I can't seem to find any reference of SpanKind in this project.

SpanKind is defined at https://github.com/census-instrumentation/opencensus-proto/blob/99162e4df59df7e6f54a8a33b80f0020627d8405/src/opencensus/proto/trace/v1/trace.proto#L88-L106

Thank you.

Any plan for cmake supporting?

Abseil-cpp or other Google project has support on cmake (using add_subdirectory). Is there any plan for opencensus-cpp?

No way to tell if StackdriverExporter::Register was successful

Ideally, the Register method for the StackdriverExporter would return a Status object. Looking at the code, this may not be trivial to implement since the registration is actually handled in a separate thread. Even so, it would still be useful to allow client code to block until registration is complete (either successful or unsuccessful) and take some action based on the result.

Use Google Kokoro CI build

Similar to grpc-java and grpc

Child Span doesn't inherit Parent's sampling decision?

The documentation says "A sampling decision from a ParentSpan is ALWAYS inherited by all its regardless of the trace configuration. This ensures continuity of traces – for example, if a parent were sampled but one of its children were not, we’d then lose parts of the trace, and vice versa if the parent weren’t sampled yet one of its children were, we wouldn’t know where the span began" https://opencensus.io/tracing/sampling/

However, looking at the code, the sampler doesn't take into consideration of parent span context.
https://github.com/census-instrumentation/opencensus-cpp/blob/master/opencensus/trace/internal/sampler.cc#L57
So I don't think the rule is enforced.

Something's wrong with AttributeValueRef or VS2017

Hello,

With Visual Studio 2017 (15.5.6) I noticed a failing test even in debug mode:

TEST(AttributeValueTest, StringValue) {
  AttributeValue attr(absl::StrCat("hello", " ", "world"));
  EXPECT_EQ(AttributeValue::Type::kString, attr.type());
  EXPECT_EQ("hello world", attr.string_value());
}

The temporary AttributeValueRef that was used to construct the AttributeValue contained a string_view which content pointer pointer into glibberish memory. When I changed the test to the following:

TEST(AttributeValueTest, StringValue) {
  std::string value = absl::StrCat("hello", " ", "world");
  AttributeValue attr(value);
  EXPECT_EQ(AttributeValue::Type::kString, attr.type());
  EXPECT_EQ("hello world", attr.string_value());
}

I noticed that before the string_view was created a temporary std::string got created, too. But instead of being kept alive until the AttributeValue was constructed, it gets deleted right before entry of AttributeValueRef ctor. Starting from that the string view points into free'd memory.

I thing the following AttributeValueRef code is to blame (or Visual Studio 2017)

  template <typename T, typename std::enable_if<std::is_constructible<
                            absl::string_view, T>::value>::type* = nullptr>
  AttributeValueRef(T string_value)
      : string_value_(absl::string_view(string_value)), type_(Type::kString) {}

If I change the signature to AttributeValueRef(const T& string_value), the error is gone. I'm no C++ language lawyer but AFAIR temporaries are best passed as const&.

Consider to generate nicer documentation like gRPC

See the their documentation here https://grpc.io/grpc/cpp/namespacegrpc.html

stats_plugin_end2end_test is flaky

There is a race condition between checking stats in the main thread and recording them in the server thread, which is leading to Travis failures. (2% flaky on my workstation, possibly more in Travis.)

Make WithContext() conditional.

Followup for #200

Add ocagent trace exporter

Agent Attribute should contain opencensus-cpp version.

This is a followup to #92 which added the Attribute.

GRPC server not recording spans when client isn't traced by opencensus

Is this expected behavior?
Because it works for opencensus-node and opencensus-python.
I figured out this happens because grpc instrumentation always looks for span context and makes a invalid context when there is no such context found from header.

gRPC example

Add an example of propagating trace information over a gRPC call.

opencensus_task should not be exposed to the user.

In opencensus/exporters/stats/stackdriver/stackdriver_exporter.h, we require the user to set a different opencensus_task for every process exporting to Stackdriver. This should be an implementation detail that the library takes care of.

examples/helloworld and website contain discrepancies on between bucket sizes and comments

Hello there, I was just going over the examples here in order to instrument a C++ app and noticed that the examples have comments talking about

opencensus-cpp/examples/helloworld/helloworld.cc

Line 68 in db9e2bb

// frontend. The view has bucket boundaries (0, 256, 65536) that will group

but the actual buckets aren't in those sizes

opencensus-cpp/examples/helloworld/helloworld.cc

Line 77 in db9e2bb

{0, 16 * kMiB, 256 * kMiB})))

which are:

0
16 * 1MiB = 16MiB
256 * 1MiB = 256MiB

The website too seems to have discrepencies https://opencensus.io/cpp.html

Please update these or if possible post them as plain values.

Add windows testing

Probably use appveyor for that.

Fallthrough return in functions returning from a switch without a default.

Our build is presently cluttered with warnings that functions that return from a switch on an enum. Absl uses a return outside the enum to suppress these warnings; I think we should either follow suit (possibly with an assert(false)) or add build options to silence the warning.

Registering the measures as recommended doesn't work properly

I find this issue while working on grpc/grpc#15853.

I register the measures following the recommendation like

::opencensus::stats::MeasureInt64 MeasureStartCount() {
  static const ::opencensus::stats::MeasureInt64 start_count =
      ::opencensus::stats::MeasureInt64::Register(
          kMeasureStartCount, kMeasureStartCount, kMeasureStartCount);
  return start_count;
}

This works fine when I build with bazel build -c dbg //test/...
But if I build with -c opt, the log shows that (1) I am attempting to register again, (2) and the data view may give me some wrong data (data fetched is 24 for a recording of value 1).

[ RUN      ] ServerLoadReportingEnd2endTest.BasicReport
D0623 13:37:14.313988446  242908 ev_posix.cc:145]            Using polling engine: epollex
D0623 13:37:14.314029956  242908 dns_resolver.cc:339]        Using native dns resolver
D0623 13:37:14.316311692  242908 ev_posix.cc:145]            Using polling engine: epollex
D0623 13:37:14.316343875  242908 dns_resolver.cc:339]        Using native dns resolver
I0623 13:37:14.316864980  242908 server_builder.cc:273]      Synchronous server. Num CQs: 1, Min pollers: 1, Max Pollers: 2, CQ timeout (msec): 10000
D0623 13:37:14.317355586  242950 load_reporter_async_service_impl.cc:78] [LRS 0x482e70] Starting a fetch-and-sample...
D0623 13:37:14.317411990  242950 load_reporter.cc:490]       [LR 0x47fe40] Starts fetching Census view data and sampling LB feedback record.
D0623 13:37:14.317422571  242950 load_reporter.cc:190]       [CVP 0x47ff80] Starts fetching Census view data.
D0623 13:37:14.317437870  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/start_count).
D0623 13:37:14.317450829  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/end_count).
D0623 13:37:14.317461294  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/bytes_sent).
D0623 13:37:14.317471768  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/bytes_received).
D0623 13:37:14.317481978  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/latency_ms).
D0623 13:37:14.317492947  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/other_call_metric_count).
D0623 13:37:14.317503341  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/other_call_metric_value).
D0623 13:37:14.317708765  242950 load_reporter_async_service_impl.cc:70] [LRS 0x482e70] Next fetch-and-sample scheduled.
D0623 13:37:14.317972062  242908 dns_resolver.cc:280]        Start resolving.
I0623 13:37:14.319444295  242948 subchannel.cc:608]          New connected subchannel at 0x7ffff0005840 for subchannel 0x7ffff0003e40
I0623 13:37:14.320059301  242908 server_load_reporting_end2end_test.cc:137] Initial request sent.
Attempt to register measure with already-registered name: name: "grpc.io/lb/start_count"; units: "grpc.io/lb/start_count"; description: "grpc.io/lb/start_count"; type: int64
I0623 13:37:15.318668918  242950 server_load_reporting_filter.cc:263] START RECORD 1
D0623 13:37:15.318829598  242950 load_reporter_async_service_impl.cc:78] [LRS 0x482e70] Starting a fetch-and-sample...
D0623 13:37:15.318846352  242950 load_reporter.cc:490]       [LR 0x47fe40] Starts fetching Census view data and sampling LB feedback record.
D0623 13:37:15.318859150  242950 load_reporter.cc:190]       [CVP 0x47ff80] Starts fetching Census view data.
D0623 13:37:15.318877675  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/start_count).
D0623 13:37:15.318889608  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/end_count).
D0623 13:37:15.318900901  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/bytes_sent).
D0623 13:37:15.318913587  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/bytes_received).
D0623 13:37:15.318924346  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/latency_ms).
D0623 13:37:15.318935930  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/other_call_metric_count).
D0623 13:37:15.318949934  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/other_call_metric_value).
D0623 13:37:15.319142178  242950 load_reporter_async_service_impl.cc:70] [LRS 0x482e70] Next fetch-and-sample scheduled.
I0623 13:37:15.319305055  242950 load_reporter_async_service_impl.cc:161] [LRS 0x482e70] Call request delivered (lb_id_: 00000000, handler: 0x7fffe8001960). Start reading the initial request...
I0623 13:37:15.319530627  242950 load_data_store.cc:256]     [PerHostStore 0x7fffe8000e80] Re-assigned orphaned store (0x7fffe800b6b0) with original LB ID of <INVALID_LBID_238dsb234890rb> to new receiver 00000000
I0623 13:37:15.319561315  242950 load_reporter.cc:376]       [LR 0x47fe40] Report stream created (host: localhost:12746, LB ID: 00000000, load key: LOAD_KEY).
I0623 13:37:15.319572060  242950 load_reporter_async_service_impl.cc:196] [LRS 0x482e70] Initial request received. Start load reporting (load balanced host: localhost:12746, interval: 5000 ms, lb_id_: 00000000, handler: 0x7fffe8001960)...
I0623 13:37:15.319682745  242950 load_reporter_async_service_impl.cc:281] [LRS 0x482e70] Sending load report (lb_id_: 00000000, handler: 0x7fffe8001960, loads count: 0)...
D0623 13:37:15.319710483  242950 load_reporter_async_service_impl.cc:247] [LRS 0x482e70] Next load report scheduled (lb_id_: 00000000, handler: 0x7fffe8001960).
I0623 13:37:15.319819823  242908 server_load_reporting_end2end_test.cc:141] Initial response received (lb_id: 00000000).
D0623 13:37:15.320015339  242908 dns_resolver.cc:280]        Start resolving.
I0623 13:37:15.320557951  242950 server_load_reporting_filter.cc:263] START RECORD 1
Attempt to register measure with already-registered name: name: "grpc.io/lb/other_call_metric"; units: "grpc.io/lb/other_call_metric"; description: "grpc.io/lb/other_call_metric"; type: double
Attempt to register measure with already-registered name: name: "grpc.io/lb/end_count"; units: "grpc.io/lb/end_count"; description: "grpc.io/lb/end_count"; type: int64
Attempt to register measure with already-registered name: name: "grpc.io/lb/bytes_sent"; units: "grpc.io/lb/bytes_sent"; description: "grpc.io/lb/bytes_sent"; type: int64
Attempt to register measure with already-registered name: name: "grpc.io/lb/bytes_received"; units: "grpc.io/lb/bytes_received"; description: "grpc.io/lb/bytes_received"; type: int64
Attempt to register measure with already-registered name: name: "grpc.io/lb/latency_ms"; units: "grpc.io/lb/latency_ms"; description: "grpc.io/lb/latency_ms"; type: int64
D0623 13:37:16.320333762  242950 load_reporter_async_service_impl.cc:78] [LRS 0x482e70] Starting a fetch-and-sample...
D0623 13:37:16.320367867  242950 load_reporter.cc:490]       [LR 0x47fe40] Starts fetching Census view data and sampling LB feedback record.
D0623 13:37:16.320377374  242950 load_reporter.cc:190]       [CVP 0x47ff80] Starts fetching Census view data.
D0623 13:37:16.320395659  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/start_count).
D0623 13:37:16.320403838  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/end_count).
D0623 13:37:16.320414540  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/bytes_sent).
D0623 13:37:16.320425090  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/bytes_received).
D0623 13:37:16.320434589  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/latency_ms).
D0623 13:37:16.320447617  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/other_call_metric_count).
D0623 13:37:16.320457110  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/other_call_metric_value).
D0623 13:37:16.320647817  242950 load_reporter_async_service_impl.cc:70] [LRS 0x482e70] Next fetch-and-sample scheduled.
D0623 13:37:17.321343602  242950 load_reporter_async_service_impl.cc:78] [LRS 0x482e70] Starting a fetch-and-sample...
D0623 13:37:17.321377090  242950 load_reporter.cc:490]       [LR 0x47fe40] Starts fetching Census view data and sampling LB feedback record.
D0623 13:37:17.321387571  242950 load_reporter.cc:190]       [CVP 0x47ff80] Starts fetching Census view data.
D0623 13:37:17.321406578  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/start_count).
D0623 13:37:17.321420644  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/end_count).
D0623 13:37:17.321436103  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/bytes_sent).
D0623 13:37:17.321460326  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/bytes_received).
D0623 13:37:17.321474007  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/latency_ms).
D0623 13:37:17.321487261  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/other_call_metric_count).
D0623 13:37:17.321501110  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/other_call_metric_value).
D0623 13:37:17.321731319  242950 load_reporter_async_service_impl.cc:70] [LRS 0x482e70] Next fetch-and-sample scheduled.
D0623 13:37:18.322311673  242950 load_reporter_async_service_impl.cc:78] [LRS 0x482e70] Starting a fetch-and-sample...
D0623 13:37:18.322347372  242950 load_reporter.cc:490]       [LR 0x47fe40] Starts fetching Census view data and sampling LB feedback record.
D0623 13:37:18.322360432  242950 load_reporter.cc:190]       [CVP 0x47ff80] Starts fetching Census view data.
D0623 13:37:18.322386536  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/start_count).
D0623 13:37:18.322398547  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/end_count).
D0623 13:37:18.322409583  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/bytes_sent).
D0623 13:37:18.322420434  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/bytes_received).
D0623 13:37:18.322431306  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/latency_ms).
D0623 13:37:18.322442936  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/other_call_metric_count).
D0623 13:37:18.322454099  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/other_call_metric_value).
I0623 13:37:18.322466875  242950 load_reporter.cc:396]       START_COUNT GOT 1
D0623 13:37:18.322608830  242950 load_data_store.cc:146]     [PerBalancerStore 0x7fffe800b6b0] Load data merged (Key: [lb_id_=<INVALID_LBID_238dsb234890rb>, lb_tag_=, user_id_=, client_ip_hex_=00000000000000000000000000000001], Value: [start_count_=1, ok_count_=0, error_count_=0, bytes_sent_=0, bytes_recv_=0, latency_ms_=0]).
I0623 13:37:18.322626284  242950 load_reporter.cc:396]       START_COUNT GOT 24
D0623 13:37:18.322645735  242950 load_data_store.cc:146]     [PerBalancerStore 0x7fffe8001760] Load data merged (Key: [lb_id_=00000000, lb_tag_=LB_TAG, user_id_=, client_ip_hex_=00000000000000000000000000000001], Value: [start_count_=24, ok_count_=0, error_count_=0, bytes_sent_=0, bytes_recv_=0, latency_ms_=0]).
D0623 13:37:18.322805456  242950 load_reporter_async_service_impl.cc:70] [LRS 0x482e70] Next fetch-and-sample scheduled.
D0623 13:37:19.323293284  242950 load_reporter_async_service_impl.cc:78] [LRS 0x482e70] Starting a fetch-and-sample...
D0623 13:37:19.323324559  242950 load_reporter.cc:490]       [LR 0x47fe40] Starts fetching Census view data and sampling LB feedback record.
D0623 13:37:19.323335550  242950 load_reporter.cc:190]       [CVP 0x47ff80] Starts fetching Census view data.
D0623 13:37:19.323355241  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/start_count).
D0623 13:37:19.323365399  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/end_count).
D0623 13:37:19.323375394  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/bytes_sent).
D0623 13:37:19.323385084  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/bytes_received).
D0623 13:37:19.323394732  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/latency_ms).
D0623 13:37:19.323404314  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/other_call_metric_count).
D0623 13:37:19.323414000  242950 load_reporter.cc:197]       [CVP 0x47ff80] Fetched view data (view: grpc.io/lb_view/other_call_metric_value).
D0623 13:37:19.323586229  242950 load_reporter_async_service_impl.cc:70] [LRS 0x482e70] Next fetch-and-sample scheduled.
I0623 13:37:20.320542192  242950 load_reporter_async_service_impl.cc:281] [LRS 0x482e70] Sending load report (lb_id_: 00000000, handler: 0x7fffe8001960, loads count: 4)...
D0623 13:37:20.320607894  242950 load_reporter_async_service_impl.cc:247] [LRS 0x482e70] Next load report scheduled (lb_id_: 00000000, handler: 0x7fffe8001960).
test/cpp/end2end/server_load_reporting_end2end_test.cc:146: Failure
      Expected: response.load().size()
      Which is: 4
To be equal to: 3
I0623 13:37:20.321082956  242950 load_reporter_async_service_impl.cc:291] [LRS 0x482e70] Shutting down the handler (lb_id_: 00000000, handler: 0x7fffe8001960, reason: OnReadDone).
D0623 13:37:20.321106585  242950 load_data_store.cc:166]     [PerBalancerStore 0x7fffe8001760] Suspended.
D0623 13:37:20.321116969  242950 load_data_store.cc:166]     [PerBalancerStore 0x7fffe800b6b0] Suspended.
I0623 13:37:20.321128956  242950 load_reporter.cc:385]       [LR 0x47fe40] Report stream closed (host: localhost:12746, LB ID: 00000000).
I0623 13:37:20.321171552  242950 load_reporter_async_service_impl.cc:334] [LRS 0x482e70] Load reporting call is notified done (handler: 0x7fffe8001960, is_cancelled: 1).
I0623 13:37:20.321293431  242950 load_reporter_async_service_impl.cc:291] [LRS 0x482e70] Shutting down the handler (lb_id_: , handler: 0x7fffe800bcc0, reason: OnRequestDelivered).
I0623 13:37:20.321577390  242950 load_reporter_async_service_impl.cc:75] [LRS 0x482e70] Fetch-and-sample is stopped.
[  FAILED  ] ServerLoadReportingEnd2endTest.BasicReport (6008 ms)

When I tried to debug this issue with logging each access to the measures, this issue disappears... 😞

Any idea or suggestion? Thanks!

Any plan to support jaeger exporter?

Jaeger is a good opentracing impl, have plan to support jaeger exporter？

confusion when using opencensus

I wanna construct a span tree in client side, first of all, I use StartSpan("name") to create a root span, and then I pass it to another function, then it can be used to create a child span as StartSpan("name", &root_span), but I don't wanna change the signature of my functions, so I need to store Span in another structure, but in its implementation, Span can't be copyed or moved. Then I think I can store SpanContext in the structure and using it to construct a new Span, but in local I can only use StartSpan which don't accept SpanContext as parameter, I'm so confused how to use opencensus effectively to achieve all these goals. Another problem is that when I pass the local Span's context to server side, then I use StartSpanWithRemoteParent to construct a new Span, but then I need to pass it to other server side functions to create more spans, now I should use StartSpanWithRemoteParent again, or I should use StartSpan, that would cause the same problem above. Hopefully, I have made it clear for you guys to understand what the problem is. Anyone knows the solution to solve these problems, thank you all.

Add support for w3c/distributed-tracing

The specs are defined here. We need to:

Add support for Tracestate
Add support for TraceContext format
Allow users to modify the Tracestate. This is not yet defined in the specs how to be implemented.

Stackdriver export metric prefix

https://github.com/census-instrumentation/opencensus-java/blob/89ecd70f633be3e74cd9d9d20934804eebf49ca6/exporters/stats/stackdriver/src/main/java/io/opencensus/exporter/stats/stackdriver/StackdriverStatsConfiguration.java#L138-L149

Similar to the Java version, it'd be quite useful for the stackdriver exporter in cpp to implement the metric name prefix as well.

how to tracing the operation between two processes

i have such a scene to trace: process_a send data to process_b, I need to track the entire process of sending data until it is accepted by process_b. the two process communicate with share memory.
now create root_span at process_a and sent the Spancontext to process_b, and create child_span with root_span's SpanContext .but this still can't track the process of delivery. i want to end the root_span in process_b, but how to do this？

Anyone knows the solution to solve this problems, thank you!！

any plan to support in multi thread?

like this link Passing Segment Context between Threads in a Multithreaded Application

As a user of opencensus-cpp, I'd want dependent library locked to specific version

Currently, opencensus-cpp uses multiple upstream libraries, including abseil, gmock, gtest, prometheus-cpp. However, when we pulled in these dependent libraries, we pulled off from master. As a user of opencensus-cpp, if I want to build opencensus-cpp library from scratch, I may get different result from time to time.

Therefore, I suggest locking down the version for both bazel and CMake for each opencensus-cpp release.

for bazel, an example is the Bazel's absl dependent:
from

    urls = ["https://github.com/abseil/abseil-cpp/archive/master.zip"],

    urls = ["https://github.com/abseil/abseil-cpp/archive/20180600.zip"],

for Cmake's absl dependent, from

	GIT_TAG "master"

	GIT_TAG "20180600"

LastValue metrics do not work with Stackdriver exporter

I'm getting this error when OpenCensus tries to export a LastValue metric:

CreateTimeSeries request failed: INVALID_ARGUMENT: Field timeSeries[0].points[0].interval.start had an invalid value of "2019-03-31T19:17:03.270968-07:00": The start time must be equal to the end time (2019-03-31T19:17:08.271862-07:00) for the gauge metric 'custom.googleapis.com/opencensus/REDACTED'.

Travis build broken by ":_*" gRPC rules.

No idea what is happening--these should be (and up til yesterday were) filtered by the "| grep -v :_" filter, and running the commands in the .travis script still works locally for me.

wiki: publish some benchmarks

Firstly, thank you for the great work on OpenCensus-CPP!

I kindly would like to ask if perhaps someone could run and publish some benchmarks on the wiki. I ask because I have been talking to companies and almost all their tech and developers would like to know what the overhead is when traced vs untraced. I believe the wiki would be a great place to publish such values and those could be updated by anyone, without having to make a pull request or distract the folks watching and maintaining this repo.

how to register filter for each channel

Now if use opencensus::RegisterGrpcPlugin(), the filter is affected for every channel.
Does is support for register grpc filter for which channel i want to use?

Add thread debugging to WithContext.

Followup on #200: add check under !NDEBUG to WithContext to make sure it's allocated and deallocated in the same thread.

wondering if there is support for make/cmake

Is there any plan for building opencensus-cpp by using make/cmake, the build tool bazel is strange to me, I've spent a lot of time to figure out how to build it correctly...But it seems still confusing...Any chance that I can use the familiar work flow to just make and build it into a library...Thanks...

CMake followups

Experimental CMake support was added in PR #238 to address issue #86. This issue is to track followup work from the PR:

What is the recommended way to gracefully shutdown?

Is there a recommended way to terminate short lived programs? I am thinking of examples, tests, or batch processes that may have completed their work but want to give OpenCensus time to complete its work before shutting down.

Implement Context

Context is an object that holds information (such as Span and TagMap) specific to an operation (such as an RPC) and propagates it implicitly to child operations, function calls, etc.

Work that needs to be done:

MSVC version supported

The current MSVC version specified is 19, but I don't think that is easy to deduce by everybody. (see https://devblogs.microsoft.com/cppblog/side-by-side-minor-version-msvc-toolsets-in-visual-studio-2017/)

MSVC 2017 has versions x..y.z where x is MSVC 2015, 2017 etc, y is the update number and z is the bugfix/patch-number.

I'm testing on MSVC 2017 15.9.7 (latest update and patch version)

OS X testing

I have a start at https://github.com/isturdy/opencensus-cpp/tree/travis-os-x. Failing to compile gRPC, predictably.

Design Tagging API

We don't currently have an API for adding/removing/replacing tags.

OpenCensus Go: https://github.com/census-instrumentation/opencensus-go/blob/ae36bd8445ff9cde752c94cf0d4f3f9dced1facc/tag/map.go

OpenCensus Java: https://github.com/census-instrumentation/opencensus-java/blob/809ed22dfe50a21426e0db20c85a28daa8ad028b/api/src/main/java/io/opencensus/tags/TagContextBuilder.java

Memory leak when not calling Span::End

In the constructor, a span adds itself to the running span store. However, it is only removed from there in End(). Since a span has no user-defined destructor, this leads to a memory leak if End() is not called on the span.

prometheus-exporter with CMake

I'd like to build with prometheus-exporter with CMake, which should be part of #244

I plan to add prometheus-exporter into CMake support on https://github.com/StevenYCChou/opencensus-cpp/tree/ycchou-add-prometheus-exporter-cmake .

I tempted to build only prometheus-cpp's core library, but when I pass

-DENABLE_PUSH=OFF -DENABLE_PULL=OFF -DENABLE_COMPRESSION=OFF -DENABLE_TESTING=OFF

through CMAKE_ARGS, it doesn't really disable testing and compression for build time. The error message is:

% cmake CMakeLists.txt                       
-- Dependency: abseil
-- Configuring done
-- Generating done
-- Build files have been written to: /Codes/opencensus-cpp/abseil-download
[100%] Built target abseil_project
-- Configuring done
-- Generating done
-- Build files have been written to: /Codes/opencensus-cpp/prometheus-download
[100%] Built target prometheus_cpp_project
-- Could NOT find GoogleBenchmark (missing: GoogleBenchmark_LIBRARY GoogleBenchmark_INCLUDE_DIR) 
CMake Error at /usr/share/cmake-3.10/Modules/FindPackageHandleStandardArgs.cmake:137 (message):
  Could NOT find ZLIB (missing: ZLIB_LIBRARY ZLIB_INCLUDE_DIR)
Call Stack (most recent call first):
  /usr/share/cmake-3.10/Modules/FindPackageHandleStandardArgs.cmake:378 (_FPHSA_FAILURE_MESSAGE)
  /usr/share/cmake-3.10/Modules/FindZLIB.cmake:112 (FIND_PACKAGE_HANDLE_STANDARD_ARGS)
  prometheus-src/pull/CMakeLists.txt:9 (find_package)


-- Configuring incomplete, errors occurred!
See also "/Codes/opencensus-cpp/CMakeFiles/CMakeOutput.log".
See also "/Codes/opencensus-cpp/CMakeFiles/CMakeError.log".

Am I doing anything wrong?

opencensus/copts.bzl build error

bazel build //examples/grpc/...:all
INFO: Invocation ID: d03519eb-4a29-427c-b063-751b4799b838
ERROR: /home/vagrant/opencensus-cpp/opencensus/copts.bzl:24:1: file '@com_google_absl//absl:copts/configure_copts.bzl' does not contain symbol 'LLVM_FLAGS'
ERROR: /home/vagrant/opencensus-cpp/opencensus/copts.bzl:37:35: Traceback (most recent call last):
File "/home/vagrant/opencensus-cpp/opencensus/copts.bzl", line 36
select({"//opencensus:llvm_compiler": (...)})
File "/home/vagrant/opencensus-cpp/opencensus/copts.bzl", line 37, in select
LLVM_FLAGS
name 'LLVM_FLAGS' is not defined
ERROR: /home/vagrant/opencensus-cpp/opencensus/copts.bzl:42:14: name 'DEFAULT_COPTS' is not defined
ERROR: error loading package 'examples/grpc': Extension file 'opencensus/copts.bzl' has errors
INFO: Elapsed time: 0.276s
INFO: 0 processes.
FAILED: Build did NOT complete successfully (0 packages loaded)
currently loading: examples/grpc

there is an error within opencensus/copts.bzl

exporters/stats/stackdriver: warnings about no return for missing default case

In this code below

opencensus-cpp/opencensus/exporters/stats/stackdriver/internal/stackdriver_utils.cc

Lines 189 to 200 in cd3f710

 switch (data.type()) { 

 case opencensus::stats::ViewData::Type::kDouble: 

 return DataToTimeSeries(view_descriptor, data.double_data(), 

 base_time_series); 

 case opencensus::stats::ViewData::Type::kInt64: 

 return DataToTimeSeries(view_descriptor, data.int_data(), 

 base_time_series); 

 case opencensus::stats::ViewData::Type::kDistribution: 

 return DataToTimeSeries(view_descriptor, data.distribution_data(), 

 base_time_series); 

 } 

 }

we can see that we have 3 different cases that all have returns but the default case doesn't

Compiling code with it with -Wreturn-type gives a bunch of warnings

INFO: From Compiling opencensus/exporters/stats/stackdriver/internal/stackdriver_utils.cc:
opencensus/exporters/stats/stackdriver/internal/stackdriver_utils.cc: In function 'google::api::MetricDescriptor::ValueType opencensus::exporters::stats::{anonymous}::GetValueType(const opencensus::stats::ViewDescriptor&)':
opencensus/exporters/stats/stackdriver/internal/stackdriver_utils.cc:86:1: warning: control reaches end of non-void function [-Wreturn-type]
 }
 ^
opencensus/exporters/stats/stackdriver/internal/stackdriver_utils.cc: In function 'std::vector<google::monitoring::v3::TimeSeries> opencensus::exporters::stats::MakeTimeSeries(const opencensus::stats::ViewDescriptor&, const opencensus::stats::ViewData&, absl::string_view)':
opencensus/exporters/stats/stackdriver/internal/stackdriver_utils.cc:200:1: warning: control reaches end of non-void function [-Wreturn-type]
 }

and in other places in the code.

In other languages such as Go that code wouldn't compile

Intern tag keys (suggestion; breaks API)

Right now, tag keys are strings, and use string semantics. This has performance implications--we need to copy/hash/check equality frequently during aggregation.

We recommend that all keys be static constants to control visibility and avoid typos. Registering tag keys in the same way as measures (e.g. static const opencensus::stats::TagKey method = opencensus::stats::RegisterTagKey("method");) would encourage these best practices and offer performance benefits throughout the library.

Implementation of install target in CMake

What about installation target in cmake?

if (gRPC_INSTALL) install(EXPORT gRPCTargets DESTINATION ${gRPC_INSTALL_CMAKEDIR} NAMESPACE gRPC:: ) endif()

Make Stackdriver BatchWriteSpans deadline configurable

I'm seeing 50%ile latency at 2.5+ seconds (<1qps) to the BatchWriteSpans endpoint for the Stackdriver exporter. Unless this is abnormal (I didn't check to see if there was anything contributing to the high latency like large request size) then the deadline should at least be override-able from client code.

Pin external repositories in WORKSPACE

http_archive rules in WORKSPACE all reference master and don't include a sha256. This leads to non-reproducible builds.

Request specific versions instead and include sha256's as described in the Bazel docs.

	switch (data.type()) {
	case opencensus::stats::ViewData::Type::kDouble:
	return DataToTimeSeries(view_descriptor, data.double_data(),
	base_time_series);
	case opencensus::stats::ViewData::Type::kInt64:
	return DataToTimeSeries(view_descriptor, data.int_data(),
	base_time_series);
	case opencensus::stats::ViewData::Type::kDistribution:
	return DataToTimeSeries(view_descriptor, data.distribution_data(),
	base_time_series);
	}
	}

census-instrumentation / opencensus-cpp Goto Github PK

opencensus-cpp's People

Contributors

Stargazers

Watchers

Forkers

opencensus-cpp's Issues

Recommend Projects

Recommend Topics

Recommend Org