-
Notifications
You must be signed in to change notification settings - Fork 890
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Decide if event.domain separately from event.name is necessary #2994
Decide if event.domain separately from event.name is necessary #2994
Comments
@open-telemetry/specs-logs-approvers FYI. |
I'm only aware of the "primary index" use case, which it seems this combined attribute would be sufficient for. Does anyone know of any other cases we need to account for? |
I think OTel should be able to define "official" prefixes within the specification repo, and we should encourage reverse FQDN for unofficial attributes (e.g. a 3rd party defining their own) |
The "group by domain" query is one such example where having one attribute doesn't work unless it is possible to unambiguously split the one attribute into the domain and name. |
My personal opinion: event domain is probably not necessary, especially for user-generated "business events" (which I honestly think is probably the #1 use-case for the event API) |
Here is another standard that uses separate fields: https://schema.ocsf.io/classes/http_activity They use 3 different fields (class_uid, category_uid, activity_id), where Interestingly then they further compose these 3 into a single |
Per conversation within the Event WG, we have decided to remove the domain field. The name field will contain namespace prefix followed by a dot and then the event name. Example:
Please reference the WG meeting notes for 2023-10-27 for discussion points. |
We, at Cisco/AppDynamics, are using |
@scheler Do you have a timeframe for which you will need to process this change? |
@scheler what do you propose in terms of migration? I think we need to decouple changes to the spec / semantic conventions from changes to APIs / instrumentation that produce event.domain. That's what's being done with the http semantic conventions: we've proceeded with breaking changes to the semantic conventions while instrumentations develop migration plans to help users with the transition. In a similar way, I think we should proceed with this change to the semantic conventions / spec and open issues with any instrumentation producing event.domain to roll out the change in an acceptable schedule. |
@jack-berg we could proceed with this change in the spec. In order to submit the changes to the semantic conventions introducing namespaced event names, I want to wait till we have a decision on how we represent event fields, so that the semantic convention PRs could include them too. Changes to the Events API should be spec'd next, and then the instrumentation changes and the migration planning could happen after that. |
As a "generalization" for this situation, I think we could include (somewhere) details that (if) the In terms of the other way around, how can a backend "determine" the |
Fixes #2994 ## Changes Merging the `domain` and `name` fields for events and modifying language to refer to the first part of the name as `namespace`
Fixes open-telemetry#2994 ## Changes Merging the `domain` and `name` fields for events and modifying language to refer to the first part of the name as `namespace`
We mentioned the possibility of merging event.domain and event.name into one attribute. Now that even.domain is no longer on the Scope having it is a separate attribute is no longer mandatory.
There are a few things to consider here:
k8s.
?The text was updated successfully, but these errors were encountered: