Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
    Tristan Sloughter
    @tsloughter
    I meant attach in the sense like using them as attributes for a span
    Marcin Grzejszczak
    @marcingrzejszczak
    You can't read from them afair though
    Tristan Sloughter
    @tsloughter
    they still exist as baggage
    Marcin Grzejszczak
    @marcingrzejszczak
    True but you can't access its contents which makes the baggage pretty useless
    Tristan Sloughter
    @tsloughter
    you can
    https://github.com/open-telemetry/opentelemetry-specification/blob/master/specification/baggage/api.md#get-all
    Marcin Grzejszczak
    @marcingrzejszczak
    You can get the baggage via the api. You can't get the baggage from a span
    Tristan Sloughter
    @tsloughter
    right, not sure why you would want to when you have the baggage. they are set as attributes on a span
    Marcin Grzejszczak
    @marcingrzejszczak
    I don't understand what you're saying . Anyways I think it's a good idea to be able to attach baggage to a span
    Tristan Sloughter
    @tsloughter
    just saying there is no need to get it from the span since it only exists there as attributes on the span. you already have the baggage if you need it. putting it on the span is a needless coupling for a system with multiple concerns within a context
    Marcin Grzejszczak
    @marcingrzejszczak
    It's not needles imo. It's a convenience method that makes sense. Brave is a mature project and putting baggage there on a span was a great idea so I'd suggest looking into it.
    Marcin Grzejszczak
    @marcingrzejszczak
    Also again i don't really care that it's in attributes if i can't retrieve it from there. It's not a good idea to always require a baggage scope to be there
    John Watson
    @jkwatson
    The thing that is hurting my head a bit is trying to think about Otel as context-centric. It's all about the context, not about spans or baggage. The only thing you propagate is the context. Period. Once you get that mental model locked in, many things become more clear. But, this is only solidifying in my head in this past week, and I've been working on this stuff for a year.
    Marcin Grzejszczak
    @marcingrzejszczak
    Fair enough so if i can attach a context and retrieve it from a span / span context then that's great. I want that context to be propagated when a child span is created. I don't want to be required to put it in scope
    Tristan Sloughter
    @tsloughter
    you can't retrieve from a span / span context. but it is attached to a context and can be retrieved from there. it is propagated with the context
    Marcin Grzejszczak
    @marcingrzejszczak
    That's what I'm suggesting - to change this approach . That you could attach the context to a span. I don't think that's unreasonable .
    I understand that this is the place to discuss such changes and since sleuth started to support otel i guess we can have a say about that ?
    Tristan Sloughter
    @tsloughter
    attaching a context to a span as in a sort of "reference" to or immutable copy of the context it was in when created? if on the other hand you mean the context were to be in the span then doing anything (like metrics, logs, baggage) would require a span. but I also don't know that we aren't just talking passed each other, I can't tell why you'd rather context in span then span in context.
    yes this is one of the places to discuss it, the main place would be the weekly Spec SIG meeting and github spec issues
    those have more ppl involved
    Andrew Hsu
    @andrewhsu

    Thanks to @bogdandrutu @jgals @justinfoote @mtwo @jmacd for joining the otel spec issue scrub 🧽 on Friday. We were able to triage the new issues in the spec repo. We also went over the open issues in oteps, and proto repos.

    Currently there are 3 issues identified as P1 that are not metrics related:

    In addition, we went over remaining high level GA planning items. We’ll discuss tracking these issues at the maintainers sig mtg.

    1 reply
    Ritick Gautam
    @riticksingh
    Can anyone know how to check if collector is receiving span ?
    Tristan Sloughter
    @tsloughter
    what are people doin gfor the description of a "parent bsaed" sampler?
    are you doing like ParentBased{TraceIdRatioBased{0.001.0}}? what about all the other options, maybe ParentBased{root=TraceIdRatioBased{0.001.0}, remoteParentSampled=AlwaysOn{}, ...}
    Tristan Sloughter
    @tsloughter
    opened an issue open-telemetry/opentelemetry-specification#1149
    Tyler Yahn
    @MrAlias
    @tsloughter in Go we pass options to the sampler to override defaults for all the different situations: https://github.com/open-telemetry/opentelemetry-go/blob/23e32b19ec227702d6f9d62b1b8aec69c075d21b/sdk/trace/sampling.go#L135-L149
    Tristan Sloughter
    @tsloughter
    this is what I'm asking abotu that isn't in the spec yet https://github.com/open-telemetry/opentelemetry-go/blob/23e32b19ec227702d6f9d62b1b8aec69c075d21b/sdk/trace/sampling.go#L254-L260
    Tyler Yahn
    @MrAlias
    Ah, gotcha :+1:
    John Watson
    @jkwatson
    [just musing out loud here] I wish that TraceState & TraceFlags were propagated via the Context, rather than hanging off the Span.
    Tristan Sloughter
    @tsloughter
    I may hav emissed it but while samplers have a "get description" function there isn't a way to get the sampler that is in use?
    and since samplers are in the SDK not API maybe there should be an API function like "TracerProvider.toString"? for use in logs that includes the sampler description
    Andrew Hsu
    @andrewhsu

    Friendly reminder we have otel spec issue scrub 🧽 mtg on Friday morning 8:30a PDT.

    We also have 2 outstanding P1 trace issues:

    Tristan Sloughter
    @tsloughter
    so invalid spans/span ctxs are used in place of having shit return like Maybe(Span), right?
    I'd been basically doing the "maybe span" but now as I'm updating how propagation works am moving to how others do it but don't think it is clear in the spec that is what is going on
    Tristan Sloughter
    @tsloughter
    hm, the Go impl generates a span id whether or not the span is recorded or sampled
    Tristan Sloughter
    @tsloughter
    python does too... I guess that is the expected way, will find it in the spec
    Tristan Sloughter
    @tsloughter
    the span creation spec doesn't say to generate a span id
    Tristan Sloughter
    @tsloughter
    I wasn't making a Span when Sampled was false and IsRecord was true, but it appears we do have to in order to call the Processors on it.
    Tristan Sloughter
    @tsloughter
    ah duh, others do since Span.set_attributes(Attributes) has to work even if Span isn't recorded and thus nothing is actually happening in that call
    Andrew Hsu
    @andrewhsu
    fyi i’ve created 11 new issues for high-level tracking of OTel GA related issues with label area:ga-tracking. i left out a priority label on purpose so it does not muck with reports of existing issue tracking of changes to the spec. essentially, i copied the line items from google sheet OpenTelemetry GA Planning to github project OpenTelemetry GA High Level Items so they can be tracked in github with assignees and status across linked github issues to other language SIGs. i’ll take a few minutes at the spec sig mtg tomorrow to give quick summary about these issues and how they will be tracked, but i’ll go over more detail in depth at the next maintainers meeting.
    Andrew Hsu
    @andrewhsu

    @tigrannajaryan @bogdandrutu from the topic of our last collector SIG mtg, i’m looking for a timeslot to triage the 148 collector issues. to keep friendly with european timezone, trying for early morning pacific time. i can suggest 2 windows of times:

    1. wed (nov 4) 7:30a - 8:00a PST; and/or
    2. thu (nov 5) 8:00a - 8:30a PST

    we’ll likely need several sessions to make a dent in the number of issues. do either/both of these times work? if so, i can setup a zoom call. else, propose an alternative time to work towards.

    11 replies
    Tristan Sloughter
    @tsloughter
    damn, could have swore everything was an hour later.. missing the spec meeting
    Ted Young
    @tedsuo
    ^ we be time shifting like doctor strange.
    Tristan Sloughter
    @tsloughter
    hah
    Jakub Malinowski
    @jtmalinowski
    hey all, I'm working on the implementation of env vars in JS as part of GA burndown, and there's one thing unclear for me, what's our (from the perspective of specification) general approach to invalid numbers in configuration, e.g. OTEL_SPAN_ATTRIBUTE_COUNT_LIMIT being set to -1? I gave myself a reasonable amount of time to try to find it in the docs, but I couldn't find anything about it.
    Tristan Sloughter
    @tsloughter
    the spec says attribute values that are arrays must be homogeneous but the protobuf actually allows it to not be.
    not sure what to do in the case of erlang, would have preferred that the export could simply fail when it couldn't encode because it finds a string in a repeated int64 ..., but it won't. since that is the case it would be better for us to just send non-homogenous arrays in attributes because otherwise we have to scan every list of attributes added to a span and check the "type" (have no types, so also a pain)
    Andrew Hsu
    @andrewhsu

    Friendly reminder we have otel spec issue scrub 🧽 mtg tomorrow, Friday morning 8:30a PST.

    Not many spec issues to triage, so if we have quorum with collector maintainers, we’ll also continue triaging collector and collector-contrib issues

    Currently, our 3 outstanding P1 trace issues:

    Tristan Sloughter
    @tsloughter
    wait what, semantic conventions are yaml now... are they published anywhere as not yaml?
    4 replies
    and http status no longer maps to any span status, right? it is only set by the user?