🟠 LOW SEV · Redis Fork Stall
🔥 Redis save fork took over 100ms — the event loop stalled, and every block check stalled with it
<|📈 graph> · <|📊 service dashboard> ·
🚨 alarm definition ·
🔕 silence
_Redis forks the process to write an RDB snapshot or rewrite the AOF. Redis is SINGLE-THREADED, so the fork blocks the event loop: while it runs, every inline proxy:block lookup and every mock-IP DNS allocation waits behind it. That is user-facing latency on the connection path, not background maintenance.
Rising fork time normally means a growing dataset or memory pressure (copy-on-write cost scales with it). Cross-check the memory headroom panel — if both are climbing, the eviction cliff is the real story and this is a symptom.
Baseline: 5ms now, 10ms max over 7d. The 100ms threshold is ~10x the observed worst case.
Panel: /d/redis-service-health 'Would a restart lose the keys?'. Ticket: CLE-514._
Grafana vmain-b1e6358