$ the-wire · showcase
SUPABASE HIBERNATION WAKES ON DASHBOARD ACCESS, ETL FIXES PORT COLLISION
By RepoJournal · Filed · About Supabase · Composed from the cited sources · methodology
3 people shipped this
Dashboard loads now automatically wake hibernated projects, while the replicator stops crashing when ClickHouse claims port 9000.
The flagship hibernation feature gets smarter this morning [1]. Users hitting the dashboard on a suspended project no longer stare at a frozen state—the system wakes it up automatically at a central checkpoint that catches most project-ref pages. For active projects, this adds zero overhead since the ref details endpoint signals whether hibernation is active [1]. Meanwhile, the ETL replicator was crashing in local dev setups because both its Prometheus metrics listener and ClickHouse's native TCP port wanted port 9000 [2]. Default configuration now maps ClickHouse to 9001 on the host side while keeping the container internal port unchanged, killing the 'address already in use' crash loop that was breaking every developer's docker-compose startup [2]. The Realtime team also tightened broadcast warmup timing to prevent edge cases in subscription synchronization [3].
Action items
- → If running local ETL dev environment, update docker-compose to use CLICKHOUSE_NATIVE_PORT=9001 supabase/etl [immediate]
- → Test hibernation wake flow on staging before rolling out to production supabase/supabase [plan]
- → Monitor Realtime broadcast latency after warmup increase supabase/realtime [monitor]
References
- [1] feat(hibernation): automated wake ↗ supabase/supabase
- [2] Move local ClickHouse native port off 9000 to avoid collision ↗ supabase/etl
- [3] chore: increase warmup on broadcast change ↗ supabase/realtime
