Queries freezing/hanging
🔢DB Connections
Hi Supabase team & community,
We're experiencing queries freezing/hanging on our Supabase Postgres instance, causing API requests to stall for 10+ seconds.
Environment
Supabase compute: Small
DB region: ap-southeast-1
App server: GCP Southeast Asia (same region)
Connection mode: Connection pooler (Supavisor)
Runtime: Bun
What's happening
Our Sentry traces show pg.pool.connection spans lasting 2–8 minutes, meaning requests are waiting that long just to acquire a connection. This stalls our API handlers for 10+ seconds and sometimes causes full timeouts under load.
What we've tried
We did a pg_dump and restored to a separate DB server with the same app and workload, things run fine there. This suggests the issue is specific to our Supabase instance rather than our application or query patterns.
Questions
Are there known limitations or bottlenecks with Supavisor on the Small compute tier that could explain this?
Are there pooler settings (pool size, max client connections, pool mode) we should tune?
Would upgrading compute likely resolve this, or is there something else to investigate first?
Happy to share Sentry traces or pg_stat_activity output if helpful. Thanks!
We're experiencing queries freezing/hanging on our Supabase Postgres instance, causing API requests to stall for 10+ seconds.
Environment
Supabase compute: Small
DB region: ap-southeast-1
App server: GCP Southeast Asia (same region)
Connection mode: Connection pooler (Supavisor)
Runtime: Bun
What's happening
Our Sentry traces show pg.pool.connection spans lasting 2–8 minutes, meaning requests are waiting that long just to acquire a connection. This stalls our API handlers for 10+ seconds and sometimes causes full timeouts under load.
What we've tried
We did a pg_dump and restored to a separate DB server with the same app and workload, things run fine there. This suggests the issue is specific to our Supabase instance rather than our application or query patterns.
Questions
Are there known limitations or bottlenecks with Supavisor on the Small compute tier that could explain this?
Are there pooler settings (pool size, max client connections, pool mode) we should tune?
Would upgrading compute likely resolve this, or is there something else to investigate first?
Happy to share Sentry traces or pg_stat_activity output if helpful. Thanks!
