CAMEL-23865: Use EWMA smoothing for throughput and fix TUI sparkline scaling by davsclaus · Pull Request #24356 · apache/camel

davsclaus · 2026-07-01T09:35:33Z

Summary

Switch the throughput metric in LoadThroughput from raw instantaneous rate to EWMA (exponentially weighted moving average) smoothing with a 1-minute decay window. This produces stable, converging throughput readings instead of oscillating 0/spike values — consistent with how LoadTriplet already computes CPU load averages.

Also fixes the TUI sparkline charts to properly display sub-1.0 msg/s rates by scaling throughput values by 100x internally.

Changes

camel-management (core):

LoadThroughput: EWMA smoothing with Math.exp(-1/60) decay, negative delta clamping, reset() stops the StopWatch
LoadThroughputTest: Awaitility-based tests verifying EWMA convergence, smoothing bounds, and reset behavior

camel-jbang-plugin-tui:

MetricsCollector: reads EWMA throughput from status JSON, stores scaled by THROUGHPUT_SCALE=100; shared niceMax() and formatThroughput() utility methods
OverviewTab / EndpointsTab: sparkline charts use scaled values with 1-2-5 Y-axis snapping; delegate to MetricsCollector for formatting

docs:

Upgrade guide entry for throughput MBean semantics change (instantaneous to EWMA)

Claude Code on behalf of Claus Ibsen

github-actions · 2026-07-01T09:36:08Z

🌟 Thank you for your contribution to the Apache Camel project! 🌟
🤖 CI automation will test this PR automatically.

🐫 Apache Camel Committers, please review the following items:

First-time contributors require MANUAL approval for the GitHub Actions to run
You can use the command /component-test (camel-)component-name1 (camel-)component-name2.. to request a test from the test bot although they are normally detected and executed by CI.
You can label PRs using skip-tests and test-dependents to fine-tune the checks executed by this PR.
Build and test logs are available in the summary page. Only Apache Camel committers have access to the summary.

⚠️ Be careful when sharing logs. Review their contents before sharing them publicly.

gnodet

Thanks for this improvement, Claus — the EWMA smoothing approach is sound and consistent with the existing LoadTriplet pattern, and the TUI sparkline scaling is a nice enhancement. The formula is correct and the overall direction is good.

A few items to address before merging (details in inline comments):

Blocking:

The new test introduces Thread.sleep() calls — project conventions require Awaitility instead.
Negative deltas are no longer guarded, which can produce persistent negative throughput via EWMA memory.

Suggestions:
3. reset() doesn't stop the watch — the first post-reset update can produce a spike that persists ~60s via EWMA.
4. niceMax and formatThroughput are duplicated between OverviewTab and EndpointsTab with divergent implementations.
5. The throughput JMX MBean attribute changed semantics (instantaneous → EWMA) — this should get an upgrade guide entry per project conventions.

This review covers project conventions and code correctness. It does not replace specialized review tools such as CodeRabbit, Sourcery, or SonarCloud.

This review was generated by an AI agent and may contain inaccuracies. Please verify all suggestions before applying.

gnodet · 2026-07-01T09:51:33Z

-            } else {
-                thp = 0;
+                // instantaneous rate in exchanges/second for this interval
+                double instantRate = (1000d / time) * delta;


The old code guarded if (delta > 0) and set thp = 0 for non-positive deltas. With EWMA, a negative instantRate (from currentReading < last, e.g. after a route restart that resets the exchange counter) bleeds into thp and takes ~60 seconds to decay back toward zero.

Consider clamping:

Suggested change

double instantRate = (1000d / time) * delta;

double instantRate = Math.max(0, (1000d / time) * delta);

Done — added Math.max(0, ...) clamping on the instantaneous rate. Also added watch.stop() in reset() so the next update() re-enters the initialization branch cleanly, avoiding post-reset spikes.

gnodet · 2026-07-01T09:51:33Z

+        t.update(total);
+        Thread.sleep(10);
+        for (int i = 0; i < 120; i++) {
+            total++;


The test introduces multiple Thread.sleep() calls. Per project conventions:

"New test code MUST NOT introduce Thread.sleep() calls. Use the Awaitility library instead."

Since LoadThroughput relies on StopWatch internally, it requires real elapsed time. A better long-term approach would be an injectable time source for fully deterministic tests. At minimum, the sleep-based timing should be replaced with Awaitility polling assertions.

Done — rewrote all tests using Awaitility's pollInterval + untilAsserted pattern. Since LoadThroughput relies on real elapsed time via StopWatch, the Awaitility polling drives the updates at 10ms intervals while the assertions verify convergence.

gnodet · 2026-07-01T09:51:33Z

+    @Test
+    public void testSmoothing() throws Exception {
+        LoadThroughput t = new LoadThroughput();
+


This assertion (> 0) would also pass with the old raw throughput calculation. Consider tightening it to verify that EWMA smoothing is actually converging toward the expected steady-state rate (e.g., within a reasonable tolerance of the true rate).

Done — tightened the convergence assertion to check throughput > 5.0 (verifying actual EWMA convergence toward the steady-state rate rather than just > 0). The smoothing test now also checks the value is bounded within a reasonable range.

gnodet · 2026-07-01T09:51:33Z

+import static org.junit.jupiter.api.Assertions.assertTrue;
+
+@DisabledOnOs(OS.AIX)
+public class LoadThroughputTest {


This test is pure Java math + Thread.sleep — no JMX interaction. The @DisabledOnOs(OS.AIX) annotation may have been cargo-culted from other management tests that interact with platform MBeans. Is there a specific AIX reason to skip this test? If not, consider removing it.

Good catch — removed. This is a pure Java math test with no JMX interaction, so the AIX exclusion was cargo-culted from other management tests.

gnodet · 2026-07-01T09:51:33Z

+     * 0.20 → 1, 1.5 → 2, 3 → 5, 7 → 10, 15 → 20, 35 → 50, 80 → 100, etc.
+     */
+    private static long niceMax(long rawMax) {
+        long s = MetricsCollector.THROUGHPUT_SCALE;


niceMax and formatThroughput are duplicated between OverviewTab and EndpointsTab with different implementations:

Here (OverviewTab): hardcoded array of 10 steps up to 1000 * THROUGHPUT_SCALE; falls back to rawMax (no rounding) beyond that.

EndpointsTab: uses a while(true) loop with 1-2-5 step sequence that scales indefinitely.

The same data could get different Y-axis ceilings depending on which chart displays it. Consider extracting both methods to a shared utility (e.g., on MetricsCollector where THROUGHPUT_SCALE already lives) with a single correct implementation.

Done — extracted both niceMax and formatThroughput to MetricsCollector as package-private static utility methods. Used the more robust EndpointsTab version (1-2-5 loop that scales indefinitely) as the single implementation. Both tabs now delegate to MetricsCollector.niceMax() and MetricsCollector.formatThroughput().

github-actions · 2026-07-01T09:54:39Z

🧪 CI tested the following changed modules:

core/camel-management
docs
dsl/camel-jbang/camel-jbang-plugin-tui

ℹ️ Dependent modules were not tested because the total number of affected modules exceeded the threshold (50). Use the test-dependents label to force testing all dependents.

Build reactor — dependencies compiled but only changed modules were tested (3 modules)

Camel :: Docs
Camel :: JBang :: Plugin :: TUI
Camel :: Management

⚙️ View full build and test results

Co-Authored-By: Claude <noreply@anthropic.com> Signed-off-by: Claus Ibsen <claus.ibsen@gmail.com>

Use THROUGHPUT_SCALE (x100) in MetricsCollector so fractional msg/s rates are preserved as longs. Apply niceMax (1-2-5 step sequence) for Y-axis scaling. Use Locale.US for consistent dot decimal formatting. Endpoint charts use showYAxis(false) until tamboui supports yAxisFormatter. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> Signed-off-by: Claus Ibsen <claus.ibsen@gmail.com>

- Clamp negative instantaneous rate to zero to prevent EWMA memory bleed - Stop StopWatch on reset() to avoid post-reset throughput spike - Replace Thread.sleep() with Awaitility in tests per project conventions - Remove unnecessary @DisabledOnOs(OS.AIX) from pure-math test - Tighten test assertions to verify EWMA convergence behavior - Extract niceMax/formatThroughput to MetricsCollector shared utility - Add upgrade guide entry for throughput MBean semantics change Co-Authored-By: Claude <noreply@anthropic.com> Signed-off-by: Claus Ibsen <claus.ibsen@gmail.com>

github-actions Bot added core dsl labels Jul 1, 2026

gnodet requested changes Jul 1, 2026

View reviewed changes

davsclaus and others added 3 commits July 1, 2026 13:11

CAMEL-23865: Use EWMA smoothing for throughput calculation

de94a3d

Co-Authored-By: Claude <noreply@anthropic.com> Signed-off-by: Claus Ibsen <claus.ibsen@gmail.com>

davsclaus force-pushed the fix/CAMEL-23865 branch from a8138de to 54b6cab Compare July 1, 2026 11:12

github-actions Bot added the docs label Jul 1, 2026

davsclaus requested a review from gnodet July 1, 2026 12:35

davsclaus merged commit 5c46803 into main Jul 1, 2026
6 checks passed

	double instantRate = (1000d / time) * delta;
	double instantRate = Math.max(0, (1000d / time) * delta);

Uh oh!

Conversation

davsclaus commented Jul 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Changes

Uh oh!

github-actions Bot commented Jul 1, 2026

Uh oh!

gnodet left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

github-actions Bot commented Jul 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

davsclaus commented Jul 1, 2026 •

edited

Loading

github-actions Bot commented Jul 1, 2026 •

edited

Loading