Create Scuba Exporter #1931
Open
Create Scuba Exporter #1931
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
meta-cla
bot
added
the
CLA Signed
|
@thomasywang has exported this pull request. If you are a Meta employee, you can view the originating Diff in D87363772. |
thomasywang
added a commit
to thomasywang/monarch-1
that referenced
this pull request
Nov 19, 2025
Summary: Stack context: ``` Our tracing subscriber has 3 layers: - File logging - Scuba - Sqlite (usually off) Although the actual Scuba logging is done in a background thread and we are using a non-blocking file writer, we still have a good chunk of work that happens for events & spans. The solution to this, is to create a `UnifiedLayer` that just sends everything into a background worker, that then delivers all traces to each `Exporter` to handle. In this diff, we will create an initial `UnifiedLayer` and incrementally move each existing layer into an `Exporter`. To test correctness, we will run both the old and unified implementations for initializing telemetry on a variety of workloads, and ensure that both are producing the same results ``` In this diff we will create an `Exporter` meant to replace `otel::tracing_layer()` (which is really just an alias for scuba). We log to two different scuba tables: monarch_tracing and monarch_executions. We will test correctness by injecting a mock scuba client that simply appends all samples it intends to log, and ensure that both the old and the unified implementations produce the same samples Differential Revision: D87363772
thomasywang
force-pushed
the
export-D87363772
branch
2 times, most recently
from
99f057a to
414b7fc
Compare
thomasywang
added a commit
to thomasywang/monarch-1
that referenced
this pull request
Nov 19, 2025
Summary: Stack context: ``` Our tracing subscriber has 3 layers: - File logging - Scuba - Sqlite (usually off) Although the actual Scuba logging is done in a background thread and we are using a non-blocking file writer, we still have a good chunk of work that happens for events & spans. The solution to this, is to create a `UnifiedLayer` that just sends everything into a background worker, that then delivers all traces to each `Exporter` to handle. In this diff, we will create an initial `UnifiedLayer` and incrementally move each existing layer into an `Exporter`. To test correctness, we will run both the old and unified implementations for initializing telemetry on a variety of workloads, and ensure that both are producing the same results ``` In this diff we will create an `Exporter` meant to replace `otel::tracing_layer()` (which is really just an alias for scuba). We log to two different scuba tables: monarch_tracing and monarch_executions. We will test correctness by injecting a mock scuba client that simply appends all samples it intends to log, and ensure that both the old and the unified implementations produce the same samples Differential Revision: D87363772
thomasywang
force-pushed
the
export-D87363772
branch
from
414b7fc to
015324b
Compare
thomasywang
added a commit
to thomasywang/monarch-1
that referenced
this pull request
Nov 19, 2025
Summary: Stack context: ``` Our tracing subscriber has 3 layers: - File logging - Scuba - Sqlite (usually off) Although the actual Scuba logging is done in a background thread and we are using a non-blocking file writer, we still have a good chunk of work that happens for events & spans. The solution to this, is to create a `UnifiedLayer` that just sends everything into a background worker, that then delivers all traces to each `Exporter` to handle. In this diff, we will create an initial `UnifiedLayer` and incrementally move each existing layer into an `Exporter`. To test correctness, we will run both the old and unified implementations for initializing telemetry on a variety of workloads, and ensure that both are producing the same results ``` In this diff we will create an `Exporter` meant to replace `otel::tracing_layer()` (which is really just an alias for scuba). We log to two different scuba tables: monarch_tracing and monarch_executions. We will test correctness by injecting a mock scuba client that simply appends all samples it intends to log, and ensure that both the old and the unified implementations produce the same samples Differential Revision: D87363772
Summary: We disallow methods relating to time to ensure that we use `hyperactor::clock`, but need to make an exception for this for telemetry because we would create a circular dependency if we tried to use `hyperactor::clock` Differential Revision: D87664116
Summary: Each process only logs to monarch_executions once at the beginning of the execution so there is no need to add a scuba client that logs to this table into our tracing subscriber Differential Revision: D87664117
Differential Revision: D87363773
Differential Revision: D87363775
Differential Revision: D87363774
Summary: Pull Request resolved: meta-pytorch#1931 Stack context: ``` Our tracing subscriber has 3 layers: - File logging - Scuba - Sqlite (usually off) Although the actual Scuba logging is done in a background thread and we are using a non-blocking file writer, we still have a good chunk of work that happens for events & spans. The solution to this, is to create a `UnifiedLayer` that just sends everything into a background worker, that then delivers all traces to each `Exporter` to handle. In this diff, we will create an initial `UnifiedLayer` and incrementally move each existing layer into an `Exporter`. To test correctness, we will run both the old and unified implementations for initializing telemetry on a variety of workloads, and ensure that both are producing the same results ``` In this diff we will create an `Exporter` meant to replace `otel::tracing_layer()` (which is really just an alias for scuba). We log to two different scuba tables: monarch_tracing and monarch_executions. We will test correctness by injecting a mock scuba client that simply appends all samples it intends to log, and ensure that both the old and the unified implementations produce the same samples Reviewed By: mariusae Differential Revision: D87363772
thomasywang
force-pushed
the
export-D87363772
branch
from
015324b to
0657ba6
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary:
Stack context:
In this diff we will create an
Exportermeant to replaceotel::tracing_layer()(which is really just an alias for scuba). We log to two different scuba tables: monarch_tracing and monarch_executions. We will test correctness by injecting a mock scuba client that simply appends all samples it intends to log, and ensure that both the old and the unified implementations produce the same samplesDifferential Revision: D87363772