Skip to content
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

Add cardinality limits to the Lightstep metrics SDK #385

Merged
merged 23 commits into from
Mar 17, 2023
Merged

Conversation

jmacd
Copy link
Member

@jmacd jmacd commented Feb 15, 2023

Description: This is a prototype for open-telemetry/opentelemetry-specification#2960.

This adds both InstrumentCardinalityLimit and AggregatorCardinalityLimit.

The instrument-level limit is meant as a circuit-breaker and the aggregator-level limit is a meant to allow views to correct overflow problems. Both of these features use otel.metrics.overflow=true to indicate the loss of attributes due to overflow. Only the synchronous instrument code path uses InstrumentCardinalityLimit in this PR, asynchronous do not, a low-priority issue to be solved separately. The problem is described in #410.

The aggregator-level limit is applied after view-configured filters are applied, so is useful to correct overflow problems.

The aggregator-level is treated as a hard-limit, when configured the aggregator will produce no more than the configured number of aggregations. The instrument-level limit is considered a circuit-breaker. For delta-temporality outputs, this should be set larger than the aggregator-level limit (considering inactivity periods) sufficiently to ensure that instrument-level records are recycled fast enough for dynamic cardinality changes.

Note there is a special case for overflow treatment in the async-delta output path (in viewstate); the cumulative overflow value is calculated so that the delta is correct. This is also low-priority, but as this is a prototype for an OTel specification, I considered it and tested it.

Link to tracking Issue: open-telemetry/opentelemetry-specification#1891

Testing: Detailed new tests in syncstate, viewstate.

Documentation: Added.

@codecov-commenter
Copy link

Codecov Report

Base: 87.84% // Head: 87.76% // Decreases project coverage by -0.09% ⚠️

Coverage data is based on head (f05909f) compared to base (1fb5140).
Patch coverage: 70.58% of modified lines in pull request are covered.

❗ Current head f05909f differs from pull request most recent head 02ef200. Consider uploading reports for the commit 02ef200 to get more accurate results

Additional details and impacted files
@@            Coverage Diff             @@
##             main     #385      +/-   ##
==========================================
- Coverage   87.84%   87.76%   -0.09%     
==========================================
  Files          74       74              
  Lines        4205     4218      +13     
==========================================
+ Hits         3694     3702       +8     
- Misses        438      442       +4     
- Partials       73       74       +1     
Impacted Files Coverage Δ
lightstep/sdk/metric/internal/syncstate/sync.go 95.02% <70.58%> (-2.60%) ⬇️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

@jmacd jmacd requested review from jaronoff97 and kristinapathak and removed request for jaronoff97 March 3, 2023 07:05
@jmacd jmacd marked this pull request as ready for review March 3, 2023 07:05
@jmacd jmacd changed the title Add a cardinality limit to the Lightstep metrics SDK Add a cardinality limits to the Lightstep metrics SDK Mar 3, 2023
@jmacd jmacd changed the title Add a cardinality limits to the Lightstep metrics SDK Add cardinality limits to the Lightstep metrics SDK Mar 3, 2023
@jmacd
Copy link
Member Author

jmacd commented Mar 3, 2023

Copy link
Contributor

@kristinapathak kristinapathak left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

looks great! one small comment typo.

lightstep/sdk/metric/sdkinstrument/performance.go Outdated Show resolved Hide resolved
Co-authored-by: Kristina Pathak <kristina.m.pathak@gmail.com>
@jmacd
Copy link
Member Author

jmacd commented Mar 8, 2023

The code reviewer in me wants to suggest this could be two PRs because the two limits are independent.
The code author in me knows it was easier to put together one PR for two related changes. 👍 👎

@jmacd jmacd merged commit 052ec8b into main Mar 17, 2023
@jmacd jmacd deleted the jmacd/cardlimit branch March 17, 2023 16:17
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

3 participants