You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Avoid including the health wait in shadow-redirect recovery duration #148
A nonblocking P3 from the independent exact-head review of PR #147 at f8ea210.
In scripts/fork/host/t3-update-guard/guard, the shadow-redirect branch only runs after HEALTH_WAIT_SECONDS expires. The known-good fallback is already serving during that wait, but outage_end "$selected" records recovery afterward, so the recovery duration can be overstated by up to about three minutes.
Follow up by recording the time the fallback is confirmed serving (or otherwise excluding the candidate health-wait interval) and add a focused regression assertion for the reported duration. This does not block merging PR #147; the outage clock now closes and no stale outage marker remains.
Source review: PR #147, head f8ea21009dc5247c6dd05bca4e02e41d4310d2f3.
Activity