You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We would like to be able to spawn Celery tasks that do not have the parent's task trace_id .
The reason for this is that sometimes it makes sense to sample the child tasks independently from the parent task ( on the server, i.e. dynamic sampling).
A typical scenario is of a task that runs seldom (once a week) and that spawns a lot of independent tasks ( e.g. one task per organisation). Having the child tasks linked by trace_id would mean that we either keep all traces from all organisations (tens of thousands) or none when we would actually need to always keep the traces for a few orgs.
Currently we can disable passing trace_id but only globally for all tasks and all errors.
This is not a usable solution since a lot of times passing the trace_id makes sense.
Solution Brainstorm
There is already a proof of concept PR opened by @jan-auer that implements the feature ( no tests though) 2331
In this proof of concept, we are passing a header in apply_async() that only disables passing the trace_id for the current call.
The text was updated successfully, but these errors were encountered:
Problem Statement
We would like to be able to spawn Celery tasks that do not have the parent's task
trace_id
.The reason for this is that sometimes it makes sense to sample the child tasks independently from the parent task ( on the server, i.e. dynamic sampling).
A typical scenario is of a task that runs seldom (once a week) and that spawns a lot of independent tasks ( e.g. one task per organisation). Having the child tasks linked by
trace_id
would mean that we either keep all traces from all organisations (tens of thousands) or none when we would actually need to always keep the traces for a few orgs.Currently we can disable passing trace_id but only globally for all tasks and all errors.
This is not a usable solution since a lot of times passing the
trace_id
makes sense.Solution Brainstorm
There is already a proof of concept PR opened by @jan-auer that implements the feature ( no tests though)
2331
In this proof of concept, we are passing a header in
apply_async()
that only disables passing thetrace_id
for the current call.The text was updated successfully, but these errors were encountered: