TD-788: Adds 'hackney_host_request_time' histogram metric

.github/workflows/erlang-checks.yml

@@ -30,7 +30,7 @@ jobs:
   run:
     name: Run checks
     needs: setup
-    uses: valitydev/erlang-workflows/.github/workflows/erlang-parallel-build.yml@v1.0.10
+    uses: valitydev/erlang-workflows/.github/workflows/erlang-parallel-build.yml@v1.0.14
     with:
       otp-version: ${{ needs.setup.outputs.otp-version }}
       rebar-version: ${{ needs.setup.outputs.rebar-version }}

src/woody_event_handler_otel.erl

@@ -25,14 +25,14 @@
 
 %% Client events
 handle_event(Event, RpcID, _Meta = #{url := Url}, _Opts) when ?IS_CLIENT_INTERNAL(Event) ->
-    with_span(otel_ctx:get_current(), mk_ref(RpcID), fun(SpanCtx) ->
-        _ = otel_span:add_event(SpanCtx, atom_to_binary(Event), #{url => Url})
+    with_existing_span(otel_ctx:get_current(), mk_ref(RpcID), fun(SpanCtx) ->
+        _ = otel_span:add_event(SpanCtx, woody_util:normalize_event_name(Event), #{url => Url})
     end);
 %% Internal error handling
 handle_event(?EV_INTERNAL_ERROR, RpcID, Meta = #{error := Error, class := Class, reason := Reason}, _Opts) ->
     Stacktrace = maps:get(stack, Meta, []),
     Details = io_lib:format("~ts: ~ts", [Error, Reason]),
-    with_span(otel_ctx:get_current(), mk_ref(RpcID), fun(SpanCtx) ->
+    with_existing_span(otel_ctx:get_current(), mk_ref(RpcID), fun(SpanCtx) ->
         _ = otel_span:record_exception(SpanCtx, genlib:define(Class, error), Details, Stacktrace, #{}),
         otel_maybe_cleanup(Meta, SpanCtx)
     end);
@@ -68,7 +68,7 @@ span_end(Ctx, Key, OnBeforeEnd) ->
             ok
     end.
 
-with_span(Ctx, Key, F) ->
+with_existing_span(Ctx, Key, F) ->
     SpanCtx = woody_util:span_stack_get(Key, Ctx, otel_tracer:current_span_ctx(Ctx)),
     _ = F(SpanCtx),
     ok.

src/woody_hackney_prometheus.erl

@@ -26,6 +26,7 @@
 -define(NB_REQUESTS, hackney_nb_requests).
 -define(TOTAL_REQUESTS, hackney_total_requests).
 -define(HOST_NB_REQUESTS, hackney_host_nb_requests).
+-define(HOST_REQUEST_TIME, hackney_host_request_time).
 -define(HOST_CONNECT_TIMEOUT, hackney_host_connect_timeout).
 -define(HOST_CONNECT_ERROR, hackney_host_connect_error).
 -define(HOST_NEW_CONNECTION, hackney_host_new_connection).
@@ -50,6 +51,13 @@ new(counter, [hackney, nb_requests]) ->
         {labels, [host]},
         {help, "Number of running requests."}
     ]),
+    true = prometheus_histogram:declare([
+        {name, ?HOST_REQUEST_TIME},
+        {registry, registry()},
+        {labels, [host]},
+        {buckets, request_time_buckets_ms()},
+        {help, "Request time."}
+    ]),
     [
         true = prometheus_counter:declare([
             {name, Name},
@@ -117,6 +125,9 @@ decrement_counter(_Name, _Value) ->
 -spec update_histogram(name(), fun(() -> ok) | number()) -> ok.
 update_histogram(_Name, Fun) when is_function(Fun, 0) ->
     Fun();
+update_histogram([hackney, Host, request_time], Value) ->
+    _ = prometheus_histogram:observe(registry(), ?HOST_REQUEST_TIME, [Host], Value),
+    ok;
 update_histogram(_Name, _Value) ->
     ok.
 
@@ -132,3 +143,16 @@ update_meter(_Name, _Value) ->
 
 registry() ->
     default.
+
+request_time_buckets_ms() ->
+    [
+        5,
+        10,
+        25,
+        50,
+        100,
+        250,
+        500,
+        1000,
+        10000
+    ].

src/woody_util.erl

@@ -16,6 +16,7 @@
 -export([span_stack_put/3]).
 -export([span_stack_get/3]).
 -export([span_stack_pop/2]).
+-export([normalize_event_name/1]).
 
 %%
 %% Internal API
@@ -95,3 +96,37 @@ span_stack_pop(Key, Context) ->
             Context1 = otel_ctx:set_value(Context, ?OTEL_SPANS_STACK, Stack1),
             {ok, SpanCtx, ParentSpanCtx, Context1}
     end.
+
+%% TODO Normalize according spec:
+%%      https://github.com/open-telemetry/semantic-conventions/blob/main/docs/general/events.md
+%%
+%%      Specifically this:
+%%
+%%          Semantic conventions MUST limit names to printable Basic Latin
+%%          characters (more precisely to U+0021 .. U+007E subset of Unicode
+%%          code points). It is recommended to further limit names to the
+%%          following Unicode code points: Latin alphabet, Numeric, Underscore,
+%%          Dot (as namespace delimiter).
+-define(EVENT_NAME_FORBIDDEN_CHARS, [<<" ">>]).
+
+-spec normalize_event_name(binary() | atom()) -> opentelemetry:event_name().
+normalize_event_name(Event) when is_atom(Event) ->
+    normalize_event_name(atom_to_binary(Event, unicode));
+normalize_event_name(Event) when is_binary(Event) ->
+    binary:replace(Event, ?EVENT_NAME_FORBIDDEN_CHARS, <<$_>>, [global]).
+
+-ifdef(TEST).
+
+-include_lib("eunit/include/eunit.hrl").
+
+-spec test() -> _.
+
+-spec normalize_event_name_test_() -> _.
+normalize_event_name_test_() ->
+    [
+        ?_assertEqual(<<"client_resolve_begin">>, normalize_event_name('client resolve begin')),
+        ?_assertEqual(<<"client_resolve/begin">>, normalize_event_name('client resolve/begin')),
+        ?_assertEqual(<<"client.resolve">>, normalize_event_name('client.resolve'))
+    ].
+
+-endif.