stacklok / toolhive / 27717871704

67%

Build Type

push

github

Committed by web-flow

Commit Message

Recover OAuth workloads from transient refresh failures (#5350)

Add AuthRetrying state for transient OAuth errors

When an OAuth refresh fails as a transient error for longer than the
in-loop short retry (~1-2 minutes at defaults), the monitor used to
mark the workload unauthenticated and exit. The workload stayed dead
even after the underlying condition cleared on its own (e.g. a brief
VPN disconnect routing requests through a different network path).
Now the monitor keeps retrying on a configurable longer cadence
(default 10 minutes) until either success or a configurable ceiling
(default 24 hours), at which point the workload is finally marked
unauthenticated.

Fixes #5349

Signed-off-by: Greg Katz <gkatz@indeed.com>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Co-authored-by: Jakub Hrozek <jakub.hrozek@posteo.se>

Coverage Stats

246 of 279 new or added lines in 8 files covered. (88.17%)

23 existing lines in 6 files now uncovered.

69459 of 103420 relevant lines covered (67.16%)

65.26 hits per line