Right-size the e2e control-plane timeouts (60s to 45s) - #84
Merged
Conversation
Follow-up to the e2e reliability fix. The 60s ceilings there were picked conservatively without data; a timeout is a ceiling not a wait (a transition that completes in 2s costs 2s), so it never slowed passing runs -- but an over-generous ceiling surfaces a genuine hang slower and can mask a real slowdown, so tighter-but-sufficient is better. The structural fix -- awaitDataPath() warming the tunnel in setUpAll -- is what actually removes the flake; these timeouts are only backstops for the one genuinely variable leg (Headscale auth latency). Sized to 45s: 50% headroom over the previously-flaky 30s, tight enough that a transition suddenly taking 40s still fails the suite. - recordUntil default 60s -> 45s (control-plane transitions) - broadcast reconnect wait 60s -> 45s - onNodeChanges first-emit 10s -> 5s (local stream; original was 2s) awaitDataPath's 90s is left as-is: it is a setup readiness gate, not a per-test wait -- it returns in seconds on success and only spends the budget if the path genuinely can't establish, where failing the whole group early would be worse. Verified at 45s: passed 36/36 clean AND 36/36 under full CPU saturation (10 hogs on 10 cores) -- a harder condition than normal CI. Total wall time unchanged (~73s), confirming the ceiling doesn't cost pass-time. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Follow-up to #83. The 60s ceilings there were conservative guesses; this right-sizes them with data.
A timeout is a ceiling, not a wait — a transition that completes in 2s costs 2s. So the value never slowed passing runs (still ~73s total). But an over-generous ceiling surfaces a genuine hang slower and can mask a real slowdown, so tighter-but-sufficient is the better call — the point you raised.
The structural fix (
awaitDataPath()warming the tunnel insetUpAll) is what removes the flake; these timeouts are only backstops for the one variable leg — Headscale auth latency. Right-sized:recordUntildefault 60s→45s, broadcast 60s→45s (50% headroom over the previously-flaky 30s; tight enough that a transition suddenly taking 40s still fails)onNodeChangesfirst-emit 10s→5s (local stream; original was 2s)awaitDataPathstays 90s — it's a setup readiness gate that returns in seconds on success, not a per-test waitVerified at 45s: 36/36 clean, and 36/36 under full CPU saturation (10 hogs/10 cores) — harder than normal CI. Total wall time unchanged. Test-only.
🤖 Generated with Claude Code