Deprecate InitGlobalTracer in favor of SetGlobalTracer #128
Conversation
yurishkuro
commented
Nov 30, 2016
•
yurishkuro
commented
- Clarify the purpose of NoopTracer
|
My only other thought would be to make the global tracer an exported |
|
I'm fine with a var. But GlobalTracer isn't backwards compatible with the current method name. DefaultTracer is an option. |
|
Meh, I thought about it some more and like what you have here. If for some reason we decide to do more than just set the (unexported) global var someday, this will be better. |
|
(i.e., LGTM) |
|
@kriskowal it's already aliased. |
| // not need to keep checking if the tracer instance is nil. | ||
| // | ||
| // For the same reason, the NoopTracer is the default "global" tracer | ||
| // (see GlobalTracer and SetGlobalTracer functions). |
There was a problem hiding this comment.
Please also note that NoopTracer is not suitable for RPC frameworks because it does not propagate baggage.
There was a problem hiding this comment.
I added a warning below. But I don't think it's fair to say that it's not suitable for RPC frameworks - it is used in TChannel, for example. Even if we took a reverse approach and built tracing on top of distributed context propagation, there'd still be a case where API for context propagation would have a no-op implementation meaning "no context propagation", and you'd have the same problem.
There was a problem hiding this comment.
Ok. It is perhaps too strong to claim that NoopTracer is suitable for RPC, since it would need to bring its own baggage propagation to the table like TChannel, just as it is too strong to claim that it is not suitable for RPC.
| @@ -1,7 +1,8 @@ | |||
| language: go | |||
|
|
|||
| go: | |||
| - 1.5 | |||
There was a problem hiding this comment.
You can't drop support for 1.5 now that you've gone 1.0 - the customer expectation is that within 1.x, there are no breaking changes. Removing this would mean we wouldn't know if 1.5 was broken within the 1.x series.
| // (etc) globals are noops. | ||
| func InitGlobalTracer(tracer Tracer) { | ||
| func SetGlobalTracer(tracer Tracer) { |
There was a problem hiding this comment.
Renaming this is a backwards breaking change. Suggest creating an alias instead, or incrementing the major version.
| @@ -25,3 +25,8 @@ func GlobalTracer() Tracer { | |||
| func StartSpan(operationName string, opts ...StartSpanOption) Span { | |||
| return globalTracer.StartSpan(operationName, opts...) | |||
| } | |||
|
|
|||
| // InitGlobalTracer is deprecated. Please use SetGlobalTracer. | |||
| func InitGlobalTracer(tracer Tracer) { | |||
There was a problem hiding this comment.
found it, thanks @bensigelman
yurishkuro
force-pushed
the
global-noop-tracer
branch
from
b541c07
to
e7527cd
Compare
yurishkuro
changed the title
| // not need to keep checking if the tracer instance is nil. | ||
| // | ||
| // For the same reason, the NoopTracer is the default "global" tracer | ||
| // (see GlobalTracer and SetGlobalTracer functions). |
There was a problem hiding this comment.
Ok. It is perhaps too strong to claim that NoopTracer is suitable for RPC, since it would need to bring its own baggage propagation to the table like TChannel, just as it is too strong to claim that it is not suitable for RPC.
`InitGlobalTracer` was deprecated for `SetGlobalTracer` in opentracing/opentracing-go#128
