enhancement(app): drive memory limiting behavior through configurable modes

[tobz]

Summary

This PR adds support for configuring memory limiting behavior, and by extension, the validation of memory bounds, through a set of configurable memory "modes."

Presently, memory limiting behavior is always on: if we detect a memory limit (either specified through configuration, or extracted from the ambient cgroups limits), we validate the bounds of the configured topology and either proceed or exit if we detect that we cannot firmly adhere to the configured bounds. Technically correct, but not particularly flexible... especially during the early stage of introducing ADP to the world.

In this PR, we address this by introducing the concept of a memory "mode": a way to specify how memory limiting and bounds validation should work. At the core, we expose three modes:

• disabled (do nothing, go forth and allocate freely)
• permissive (evaluate bounds and warn if we may exceed them)
• strict (evaluate bounds and exit if we may exceed them)

We default to disabled to ensure that users have an equivalent experience when using ADP as they're used to with the Datadog Agent. Permissive and strict are designed to allow for gradually opting in to memory limiting behavior: allowing things like the global memory limiter to actively influence component behavior, dialing in configuration settings to explore the resulting memory bounds, eventually culminating in (hopefully) switching to strict mode to lock in their tuning.

The global memory limiter is also left as-is, in terms of being configurable but defaulting to enabled. However, in the "disabled" mode, we simply set the memory limiter's grant to an (essentially) unbounded value. This was simply the expedient design choice but I'm thinking I may forcefully disable it when in the "disabled" memory mode just to be a little cleaner.

Change Type

• Bug fix
• New feature
• Non-functional (chore, refactoring, docs)
• Performance

How did you test this PR?

Existing tests, and five new integration tests that exercise the different memory modes.

References

DADP-2

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f392a77413

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Return a no-op limiter when disabled

When memory_mode is left at its new default disabled and enable_global_limiter is also left at its default true, returning Some(MemoryGrant::unbounded()) still drives the later MemoryLimiter::new(grant) path. That means a mode documented as disabled now spawns the limiter checker and can fail startup with Memory statistics cannot be gathered on this system on platforms where RSS stats are unavailable, whereas the previous no-limit path returned MemoryLimiter::noop() without touching process-memory APIs. The disabled branch should avoid creating an active limiter grant.

Useful? React with 👍 / 👎.

[pr-commenter]

Regression Detector (Agent Data Plane)

Regression Detector Results

Run ID: 70e46712-32f4-44ea-bb1d-d7d64ca0e432

Baseline: 0f0001b
Comparison: 2490a72
Diff

Optimization Goals: ✅ No significant changes detected

Experiments ignored for regressions

Regressions in experiments with settings containing erratic: true are ignored.

perf	experiment	goal	Δ mean %	Δ mean % CI	trials	links
➖	otlp_ingest_logs_5mb_memory	memory utilization	+1.65	[+1.44, +1.87]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_logs_5mb_cpu	% cpu utilization	+0.80	[-3.68, +5.28]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_logs_5mb_throughput	ingress throughput	-0.00	[-0.12, +0.12]	1	(metrics) (profiles) (logs)

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI	trials	links
➖	dsd_uds_1mb_3k_contexts_cpu	% cpu utilization	+6.76	[-47.22, +60.73]	1	(metrics) (profiles) (logs)
➖	dsd_uds_10mb_3k_contexts_cpu	% cpu utilization	+2.59	[-27.22, +32.41]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_logs_5mb_memory	memory utilization	+1.65	[+1.44, +1.87]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_traces_5mb_cpu	% cpu utilization	+1.41	[-0.60, +3.42]	1	(metrics) (profiles) (logs)
➖	dsd_uds_512kb_3k_contexts_cpu	% cpu utilization	+1.35	[-54.08, +56.78]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_traces_ottl_filtering_5mb_throughput	ingress throughput	+1.16	[+1.06, +1.26]	1	(metrics) (profiles) (logs)
➖	dsd_uds_500mb_3k_contexts_throughput	ingress throughput	+1.00	[+0.88, +1.12]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_metrics_5mb_memory	memory utilization	+0.82	[+0.62, +1.02]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_logs_5mb_cpu	% cpu utilization	+0.80	[-3.68, +5.28]	1	(metrics) (profiles) (logs)
➖	dsd_uds_500mb_3k_contexts_cpu	% cpu utilization	+0.52	[-0.98, +2.02]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_traces_ottl_transform_5mb_memory	memory utilization	+0.24	[+0.08, +0.40]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_traces_ottl_filtering_5mb_memory	memory utilization	+0.21	[-0.02, +0.45]	1	(metrics) (profiles) (logs)
➖	dsd_uds_500mb_3k_contexts_memory	memory utilization	+0.21	[+0.06, +0.35]	1	(metrics) (profiles) (logs)
➖	quality_gates_rss_dsd_ultraheavy	memory utilization	+0.21	[+0.07, +0.34]	1	(metrics) (profiles) (logs)
➖	dsd_uds_100mb_3k_contexts_cpu	% cpu utilization	+0.19	[-5.61, +5.99]	1	(metrics) (profiles) (logs)
➖	quality_gates_rss_dsd_medium	memory utilization	+0.16	[-0.01, +0.33]	1	(metrics) (profiles) (logs)
➖	dsd_uds_512kb_3k_contexts_memory	memory utilization	+0.13	[-0.01, +0.27]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_traces_ottl_filtering_5mb_cpu	% cpu utilization	+0.07	[-2.09, +2.23]	1	(metrics) (profiles) (logs)
➖	quality_gates_rss_idle	memory utilization	+0.07	[+0.03, +0.11]	1	(metrics) (profiles) (logs)
➖	dsd_uds_10mb_3k_contexts_throughput	ingress throughput	+0.02	[-0.17, +0.21]	1	(metrics) (profiles) (logs)
➖	dsd_uds_100mb_3k_contexts_memory	memory utilization	+0.01	[-0.14, +0.16]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_metrics_5mb_throughput	ingress throughput	+0.00	[-0.15, +0.16]	1	(metrics) (profiles) (logs)
➖	dsd_uds_512kb_3k_contexts_throughput	ingress throughput	+0.00	[-0.05, +0.05]	1	(metrics) (profiles) (logs)
➖	dsd_uds_1mb_3k_contexts_throughput	ingress throughput	+0.00	[-0.06, +0.06]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_logs_5mb_throughput	ingress throughput	-0.00	[-0.12, +0.12]	1	(metrics) (profiles) (logs)
➖	dsd_uds_100mb_3k_contexts_throughput	ingress throughput	-0.01	[-0.05, +0.02]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_traces_5mb_memory	memory utilization	-0.02	[-0.18, +0.14]	1	(metrics) (profiles) (logs)
➖	dsd_uds_10mb_3k_contexts_memory	memory utilization	-0.08	[-0.23, +0.08]	1	(metrics) (profiles) (logs)
➖	quality_gates_rss_dsd_heavy	memory utilization	-0.10	[-0.23, +0.03]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_traces_5mb_throughput	ingress throughput	-0.19	[-0.26, -0.12]	1	(metrics) (profiles) (logs)
➖	dsd_uds_1mb_3k_contexts_memory	memory utilization	-0.21	[-0.35, -0.07]	1	(metrics) (profiles) (logs)
➖	quality_gates_rss_dsd_low	memory utilization	-0.24	[-0.40, -0.08]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_traces_ottl_transform_5mb_throughput	ingress throughput	-0.32	[-0.39, -0.24]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_metrics_5mb_cpu	% cpu utilization	-0.75	[-6.50, +5.01]	1	(metrics) (profiles) (logs)
➖	otlp_ingest_traces_ottl_transform_5mb_cpu	% cpu utilization	-1.24	[-3.25, +0.78]	1	(metrics) (profiles) (logs)

Bounds Checks: ✅ Passed

perf	experiment	bounds_check_name	replicates_passed	observed_value	links
✅	quality_gates_rss_dsd_heavy	memory_usage	10/10	123.60MiB ≤ 140MiB	(metrics) (profiles) (logs)
✅	quality_gates_rss_dsd_low	memory_usage	10/10	40.71MiB ≤ 50MiB	(metrics) (profiles) (logs)
✅	quality_gates_rss_dsd_medium	memory_usage	10/10	60.65MiB ≤ 75MiB	(metrics) (profiles) (logs)
✅	quality_gates_rss_dsd_ultraheavy	memory_usage	10/10	177.42MiB ≤ 200MiB	(metrics) (profiles) (logs)
✅	quality_gates_rss_idle	memory_usage	10/10	27.64MiB ≤ 40MiB	(metrics) (profiles) (logs)

Explanation

Confidence level: 90.00%
Effect size tolerance: |Δ mean %| ≥ 5.00%

Performance changes are noted in the perf column of each table:

• ✅ = significantly better comparison variant performance
• ❌ = significantly worse comparison variant performance
• ➖ = no significant change in performance

A regression test is an A/B test of target performance in a repeatable rig, where "performance" is measured as "comparison variant minus baseline variant" for an optimization goal (e.g., ingress throughput). Due to intrinsic variability in measuring that goal, we can only estimate its mean value for each experiment; we report uncertainty in that value as a 90.00% confidence interval denoted "Δ mean % CI".

For each experiment, we decide whether a change in performance is a "regression" -- a change worth investigating further -- if all of the following criteria are true:

1. Its estimated |Δ mean %| ≥ 5.00%, indicating the change is big enough to merit a closer look.

2. Its 90.00% confidence interval "Δ mean % CI" does not contain zero, indicating that if our statistical model is accurate, there is at least a 90.00% chance there is a difference in performance between baseline and comparison variants.

3. Its configuration does not mark it "erratic".

[lucastemb]

I know you mentioned wanting to go over the doc comments, but the code looks good to me 👍

[pr-commenter]

Binary Size Analysis (Agent Data Plane)

Target: 0f0001b (baseline) vs 2490a72 (comparison) diff
Analysis Type: Stripped binaries (debug symbols excluded)
Baseline Size: 37.08 MiB
Comparison Size: 37.10 MiB
Size Change: +20.16 KiB (+0.05%)
Pass/Fail Threshold: +5%
Result: PASSED ✅

Changes by Module

Module	File Size	Symbols
tracing	-8.14 KiB	3
figment	+7.92 KiB	15
hashbrown	+7.54 KiB	33
core	-6.60 KiB	644
bytes	+6.48 KiB	13
h2	+5.89 KiB	105
anyhow	+5.71 KiB	34
serde_json	-4.44 KiB	7
saluki_app::logging::build_output_stack	+4.04 KiB	2
rustls	+3.82 KiB	2
http_body_util	-3.28 KiB	1
tracing_appender	-3.28 KiB	3
crossbeam_channel	-2.96 KiB	10
tracing_subscriber	+2.88 KiB	26
saluki_app::logging::syslog	-2.56 KiB	4
[sections]	+2.25 KiB	7
serde_html_form	-2.25 KiB	4
tower	+1.91 KiB	17
tokio	-1.89 KiB	55
[Unmapped]	+1.74 KiB	1

Detailed Symbol Changes

    FILE SIZE        VM SIZE    
 --------------  -------------- 
  +0.9% +34.0Ki  +1.0% +28.6Ki    [2656 Others]
  [NEW] +26.4Ki  [NEW] +26.2Ki    _<figment::value::de::ConfiguredValueDe<I> as serde_core::de::Deserializer>::deserialize_struct::h025543ea3fec2e45
  [NEW] +8.55Ki  [NEW] +8.45Ki    h2::proto::connection::Connection<T,P,B>::poll::h25f96fc3015fa649
  [NEW] +8.34Ki  [NEW] +8.24Ki    h2::proto::connection::Connection<T,P,B>::poll::hf7363029696bd16e
  [NEW] +6.17Ki  [NEW] +6.04Ki    _<core::future::poll_fn::PollFn<F> as core::future::future::Future>::poll::hf8fe7ccdd5dcfb77
  [NEW] +5.14Ki  [NEW] +5.04Ki    saluki_app::memory::MemoryBoundsConfiguration::try_from_config::h3277a01cd29194c5
  [NEW] +4.78Ki  [NEW] +4.62Ki    _<figment::value::de::ConfiguredValueDe<I> as serde_core::de::Deserializer>::deserialize_any::h2a207968b7c35151
   +95% +4.25Ki   +96% +4.25Ki    saluki_app::logging::build_output_stack::h8c73f95efff6aca6
   +42% +3.80Ki   +43% +3.80Ki    _<rustls::error::Error as core::clone::Clone>::clone::h5d268dcb651f6988
  [NEW] +3.79Ki  [NEW] +3.70Ki    saluki_app::memory::initialize_memory_bounds::h838fd3bddba7987d
  [NEW] +3.43Ki  [NEW] +3.31Ki    _<bytes::bytes_mut::BytesMut as bytes::buf::buf_mut::BufMut>::put::hb1205e20738e6a41
  [NEW] +2.95Ki  [NEW] +2.84Ki    h2::proto::streams::streams::DynStreams<B>::recv_push_promise::h98b1756e4ccce078
 -74.0% -3.25Ki -75.7% -3.25Ki    h2::proto::streams::streams::StreamRef<B>::send_data::h6f07527fb0901ff4
 -69.2% -3.28Ki -71.8% -3.28Ki    _<http_body_util::util::BufList<T> as bytes::buf::buf_impl::Buf>::copy_to_bytes::hc11e5a80535d90e3
  [DEL] -4.54Ki  [DEL] -4.44Ki    saluki_app::memory::MemoryBoundsConfiguration::try_from_config::h938a0705c0d1d5a1
  [DEL] -4.95Ki  [DEL] -4.76Ki    serde_json::value::de::_<impl serde_core::de::Deserialize for serde_json::value::Value>::deserialize::hfd46eabc0227358d
 -48.8% -7.02Ki -49.3% -7.02Ki    _<tracing::instrument::Instrumented<T> as core::future::future::Future>::poll::hf6beb522713a97cb
  [DEL] -14.4Ki  [DEL] -14.3Ki    h2::server::Connection<T,B>::poll_closed::hb373424f939b3e4c
 -72.2% -14.7Ki -72.6% -14.7Ki    h2::proto::connection::DynConnection<B>::recv_frame::h18bf0080f259e42b
  [DEL] -15.3Ki  [DEL] -15.2Ki    h2::server::Connection<T,B>::poll_closed::h1236950a7f1f4ba4
  [DEL] -23.9Ki  [DEL] -23.7Ki    _<figment::value::de::ConfiguredValueDe<I> as serde_core::de::Deserializer>::deserialize_struct::hcbf3c952b886a73f
  +0.1% +20.2Ki  +0.0% +14.3Ki    TOTAL