Stop crying wolf on transient network failures (1.7.18)#45
Merged
Conversation
A scheduled sync fails most often because the network was briefly gone (the Mac asleep, Wi-Fi flapping) when the 4-hourly timer fired. Those runs self-heal on the next attempt, but each one fired a "eufy-sync failed" notification, which trains the user to ignore the notifications that matter. Headless runs whose only failures are transient network errors (DNS, TLS handshake timeouts, connection resets, "all connection attempts failed") now stay silent until three land in a row, then fire one notification that the network has been down a while and syncs are waiting. Any success resets the streak. Real failures (bad password, revoked token, rate limit, profile ambiguity) still notify immediately, every time, and interactive runs are never silenced. Classification is by message substring in a new failure_notify module, tuned to catch httpx/requests/stdlib network shapes while leaving persistent TLS problems (certificate verify failed, wrong version number - a bad clock, stale CA bundle, or intercepting proxy) as real failures that notify. The streak counter is a small state file; corrupt contents reset it rather than crash the run. Co-Authored-By: Claude Fable 5 <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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Why
The 4-hourly Launch Agent fails most often because the network was briefly unavailable (the Mac asleep, Wi-Fi flapping) when the timer fired. These runs self-heal on the next attempt, but each one fired a "eufy-sync failed" notification. Over this machine's log lifetime that is 15 false alarms, which trains the user to ignore the notifications that actually matter (a bad password, a revoked token).
What
Headless runs whose failures are all transient network errors now stay silent until three land in a row, then fire one notification saying the network has been down a while and syncs are waiting. Any successful run resets the streak.
certificate verify failed/wrong version number(a wrong system clock, a stale CA bundle, or a TLS-intercepting proxy) are real and notify.failure_notifymodule, tuned to httpx / requests / stdlib network shapes. The streak counter is a small state file under~/.garmin-sync; corrupt contents reset it rather than crash the run.Design decision
Escalate-after-3 (silent for isolated blips, one heads-up if the network is genuinely down ~12h) was chosen over always-soft-notify and fully-silent.
Testing
264 tests pass, run hermetically. New tests pin the classification (including the SSL-cert cases that must not be silenced and the httpx shapes that must be), the escalate-once-at-threshold logic, streak reset on success, corrupt-file robustness, and the headless-vs-interactive behavior end to end through
main(). Reviewed adversarially for the one dangerous failure mode (a real failure being silently suppressed); confirmed clean.🤖 Generated with Claude Code