I’ve created a PR to bump the version of the specification to v1.0. I see one or two remaining issues officially marked as required for tracing, but I want this PR to exist to raise awareness: open-telemetry/opentelemetry-specification#1372
I'm not finding this defined in the spec. do implementations use the OTEL_LOG_LEVEL to set the regular logger log level for all modules/classes/etc within the SDK
should it be exposed so that third party libs, like exporters would use the same level
in JS we export a console logger that respects this var if it is constructed without a level specified in code. each component accepts an optional logger in the constructor. For classes like span which are not created directly, they get the logger from the component that created them (in this case the tracer).
In java, we don't support OTEL_LOG_LEVEL because we want to use normal java logging APIs.
yea, we are using normal Erlang logging APIs but I was going to have OTEL_LOG_LEVEL just set the log level for the otel SDK modules
but then realized I wasn't sure if this was the proper interpretation
if it was I would have expected Java to have it also set the log level for the normal java logging APIs
but yea, a user could just use the normal erlang logger api to set the log level on the opentelemetry sdk application
Hi, I have a distributed stream processing system, and I would like to identify a suitable technology to track the provenance of the output records. For example, given the id of an output record, I would like to obtain a report which details the ids of all the input records that went into computing the results of the output record. Since the stream processing system has a number of layers, this fine-grained provenance report will be a DAG structure with the root parent as the output record, and the descendants as the input records and all the intermediate records. I'm exploring whether OpenTelemetry can fulfill this requirement. It looks like Trace and Span create such a DAG but its purpose seems different, as the DAG structure seems to be inverted, i.e. the root parent acts as the input (e.g. the external client request) and the descendants act as the outputs (e.g. the processes that are triggered by the external client request). Does OpenTelemetry address this need?
anyone else read "reference implementation" as "don't use this in production"? maybe its just me, but when I see that language, like I just saw it is used for the Java SDK it makes me think its not what you should actually use. Like Weston is the "reference implementatin" of Wayland and no one uses it for a daily driver
@bogdandrutu Following up on version numbers for the spec. You mentioned that if breaking changes to experimental components resulted in rapidly bumping the major version number (which semver would imply) then we would have a failure of some kind. I want to know more details about why this would be an issue. I think the failure of being at v12.0 of the spec by the time metrics was released– and I’m not really sure that it is a failure based on the version alone – would be that we are thrashing in the experimental work. So maybe this relates to what @jsuerth has pointed out – even in experimental work we should start taking care to avoid dribbling out breaking changes. We should take care to avoid breaking changes, and when we do make them, focus on batching the breaking changes. At the same time, figure out how we can be prototyping work on metrics while still offering a somewhat stable experience for beta users. Because otherwise, we will have angry users, regardless of the version of the spec.
I'm not super familiar with the log data model, but this is the purpose of the Resource in OTel
@wdengw there is no notion of Application in OpenTelemetry. It is not clear how Application is different from the Service. If you believe Application should be a concept recognized by OpenTelemetry then it may be worth making a proposal to add semantic conventions to OpenTelemetry. The logs will just use whatever the OpenTelemetry conventions say, there is nothing special about logs from this perspective.
in the spec for process resource attributes it says, "On Windows, can be set to the base name of GetProcessImageFileNameW."
does anyone here know what that means? is that a command that can be run in Powershell or something?
Should we move this room to cncf slack?
Seeking advice on open-telemetry/opentelemetry-go#1122 -- are SDKs required to support OTLP/HTTP to conform to Spec 1.0, or is it enough to support OTLP/gRPC (and plan to add HTTP later on)?
I think that's marked as "optional" in the compliance matrix. Also... head over to CNCF slack. There's a brand new #otel-specification channel
Thanks! I hesitated to post there because it was empty :)
Can anyone tell me how to use elastic search as a storage for zipkin+ OTEL in node js
Hi guys I want to push metrics from my Kinesis Data analytics to OTLP collector but I found that python and other opentelemetry packages have marked metrics api as alpha and not stable .How should I go forward any advise ? The problem statement is get meaningful metrics from KDA and export them to Prometheus .
How can I send the otel collection agents logs to a file when running in a docker container? I looked into the file exporter but that is much to verbose for what im looking for. I am primarly interested in the otel collectors startup logs. Is the best/only way to capture the stdout logs to use a docker log driver?