OXY-1514: only track sampled spans and through a setting track all spans. #107
Conversation
This file contains 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
We're seeing high lock contention when removing spans from the `LiveReferenceSet`. Currently, we sample all spans that are created, even if those spans are unsampled. The purpose of this is to capture graceful restart timeout issues which sometimes occur in our services. By fetching the list of all spans in memory when a graceful restart hangs, we can figure out what is preventing the restart from occuring within its timeout. In this commit we only track _sampled_ spans. Most services which use foundations sample at a 1% rate, meaning we effectively remove 99% of locks. This allows light debugability in all running services, at any time you can get a list of traced spans. Although those spans should end up in jaeger/oltp at some point[^1]. To make sure we can still determine the root cause of graceful restart issues, we add a setting which forces all spans to be sampled (what happened previously). [^1]: to be fair, sometimes some traces will remain in memory forever and never be dropped, so even at 1% this can be useful.
fisherdarling
force-pushed
the
fisher/OXY-1514
branch
from
73e5d9c to
d0d0d02
Compare
fisherdarling
commented
Jan 30, 2025
| @@ -26,6 +26,13 @@ pub struct TracingSettings { | |||
|
|
|||
| /// The strategy used to sample traces. | |||
| pub sampling_strategy: SamplingStrategy, | |||
| /// Enable liveness tracking of all generated spans. Even if the spans are | |||
| /// unsampled. This can be useful for debugging potential hangs cause by | |||
There was a problem hiding this comment.
Suggested change
| /// unsampled. This can be useful for debugging potential hangs cause by | |
| /// unsampled. This can be useful for debugging potential hangs caused by |
| @@ -26,6 +26,13 @@ pub struct TracingSettings { | |||
|
|
|||
| /// The strategy used to sample traces. | |||
| pub sampling_strategy: SamplingStrategy, | |||
| /// Enable liveness tracking of all generated spans. Even if the spans are | |||
| /// unsampled. This can be useful for debugging potential hangs cause by | |||
| /// some objects remaining in memory. The default value is false, meaning | |||
There was a problem hiding this comment.
Suggested change
| /// some objects remaining in memory. The default value is false, meaning | |
| /// some objects remaining in memory. The default value is `false`, meaning |
| pub(crate) use live_reference_set::LiveReferenceHandle; | ||
|
|
||
| use crate::telemetry::tracing::internal::SharedSpanHandle; | ||
| // pub(crate) type SharedSpanHandle = Arc<SharedSpanHandle>; |
There was a problem hiding this comment.
Suggested change
| // pub(crate) type SharedSpanHandle = Arc<SharedSpanHandle>; |
| pub(crate) struct SharedSpanInner(SharedSpanHandle); | ||
| pub(crate) enum SharedSpanHandle { | ||
| Tracked(Arc<LiveReferenceHandle<Arc<RwLock<Span>>>>), | ||
| Unsampled(Arc<RwLock<Span>>), |
There was a problem hiding this comment.
bikeshed?
Suggested change
| Unsampled(Arc<RwLock<Span>>), | |
| Untracked(Arc<RwLock<Span>>), |
fisherdarling
force-pushed
the
fisher/OXY-1514
branch
from
d0d0d02 to
84099ac
Compare
inikulin
reviewed
Feb 3, 2025
| /// only sampled spans are tracked. | ||
| /// | ||
| /// To get a json dump of the currently active spans, query: `/debug/traces` | ||
| pub track_all_spans: bool, |
There was a problem hiding this comment.
I think we need to make the whole liveness tracking functionality, i.e. introduce LivenessTrackingSettings with enabled field and also put this field in there too.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
We're seeing high lock contention when removing spans from the
LiveReferenceSet. Currently,we sample all spans that are created, even if those spans are unsampled. The purpose of this
is to capture graceful restart timeout issues which sometimes occur in our services. By fetching
the list of all spans in memory when a graceful restart hangs, we can figure out what is preventing
the restart from occuring within its timeout.
In this commit we only track sampled spans. Most services which use foundations sample at a 1% rate,
meaning we effectively remove 99% of locks. This allows light debugability in all running services, at
any time you can get a list of traced spans. Although those spans should end up in jaeger/oltp at some
point1. To make sure we can still determine the root cause of graceful restart issues, we add a setting
which forces all spans to be sampled (what happened previously).
Footnotes
to be fair, sometimes some traces will remain in memory forever and never be dropped, so even
at 1% this can be useful. ↩