Files
Dave HortonandClaude Fable 5 560c18ef68 feat: report call count to redis for server-wide drain coordination
A server may run several sbc-inbound and sbc-outbound processes.  When
one of the sbc-inbound processes handles an AWS autoscale drain it must
complete the terminating lifecycle hook only when the whole server is
idle, not just its own process — otherwise the instance is terminated
while sibling processes still have calls in progress.

Each process now writes its own call count to a per-pid redis key
(120s expiry, refreshed every 15s from activeCallIds.size) and
registers that key in a per-host set, which the draining process
enumerates and sums.  See the companion change in sbc-inbound.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-30 15:47:25 -04:00
..
2024-11-19 09:37:53 -05:00
2025-10-21 07:52:29 -04:00
2026-05-01 08:43:13 -04:00