Span data is stored and manipulated in an ETS table. Span context must be tracked in order to create child span's, propagate span context across process or node boundaries and for the library to which span data to manipulate in the context it is called.
opencensus
provides two methods for tracking this context, the process dictionary and a variable holding a ctx
record.
With ocp
the span context is tracked in the current process`s process dictionary.
some_fun() ->
ocp:with_child_span(<<"some_fun/0">>,
fun() ->
... body ..
end).
The parse transform provides an attribute to decorate functions with that will start a span, wrap the contents in a try
and finish the span in an after
clause. Add the parse transform to the compile opts in rebar.config
:
{erl_opts, [{parse_transform, oc_transform}]}.
And use:
-span([]).
function_to_trace() ->
...
SpanCtx = ocp:current_span(), %% ocp works in a decorated function too
...
Since the tranformed functions use the process dictionary to store the context you can interact with the current span the same way as you do with ocp
, covered in the previous section.
ctx
is a generic context library for Erlang. OpenCensus provides the option to use it in place of the process dictionary for tracking the span context.
In this example a function is passed a ctx
variable Ctx
that some instrumented library could have set the span context based on the incoming metadata of a request, like HTTP headers. The oc_trace:new_span
function will check Ctx
for a span context and create a child span of that span context if it exists, otherwise a root span will be created. We can pass the span context to another function, we could also createa a new ctx
to pass (oc_trace:with_span(Ctx, SpanCtx)
), to be further updated or have new children created:
handler(Ctx, NextHandler) ->
SpanCtx = oc_trace:with_child_span(Ctx, <<"span-name">>),
try
oc_trace:put_attribute(<<"key">>, <<"value">>, SpanCtx),
{Code, Message} = NextHandler(SpanCtx),
oc_trace:set_status(Code, Message, SpanCtx)
after
oc_trace:finish_span(SpanCtx)
end.
The module oc_span
has the functional span data manipulation functions, meaning ETS is not involved. Most users will not need this, but for potential alternative span data stores or context trackers they are necessary.
A span has a map of attributes providing details about the span. The key is a binary string and the value of the attribute can be a binary string, integer, or boolean.
Span1 = oc_trace:put_attribute(<<"/instance_id">>, <<"my-instance">>, SpanCtx),
A time event is a timestamped annotation with user-supplied key-value pairs or a message event to represent a message (not specificly an Erlang message) sent to or received from another span.
The message_event
consists of a type, identifier and size of the message. Id
is an identifier for the event's message that can be used to match SENT
and RECEIVED
message_event
s. For example, this field could represent a sequence ID for a streaming RPC. It is recommended to be unique within a Span. If CompressedSize
is 0
it is assumed to be the same as UncompressedSize
.
Event = opencensus:message_event(?MESSAGE_EVENT_TYPE_SENT, Id, UncompressedSize, CompressedSize)
oc_trace:add_time_event(Event, SpanCtx),
Links are useful in cases like a job queue. A job is created with a span context and when run wants to report a new span. The job isn't a direct child of the span that inserted it into the queue, but it is related. The job creates a link to the span that created it.
SpanCtx = oc_trace:with_child_span(Ctx, <<"running job">>),
Link = oc_trace:link(?LINK_TYPE_PARENT_LINKED_SPAN, TraceId, ParentSpanId, #{}),
oc_trace:add_link(Link, SpanCtx),
... run job ...
oc_trace:finish_span(SpanCtx).