FATAL: the database system is in recovery mode

[sbutlerjr]

I get this error, "FATAL: the database system is in recovery mode" sometimes multiple times per day. It is usually followed by "FATAL: the database system is not accepting connections. Hot standby mode is disabled."

It happens quickly and then is available usually in under a second or two, but its started to cause issues in my application. Looking at stats I shared below, i'm barely putting any stress on the system.

I've very experienced with Postgres and would normally investigate the logs, but Supabase just doesn't show that level of detail. Are there other logs I'm missing?

Looking at stats for last 24 hrs, which include multiple of above errors:
Compute size: MICRO (1GB memory, 2-core ARM CPU)
Database Size: 0.09 GB (8gb provisioned)
Max CPU Usage: 3.47%
Memory Usage: 56.03%
Disk IOps write 2
Disk IOps read: 0
Pooler to database connections: 13 (sometimes goes up to 30)
Client to Dedicated Pooler connections: 1
API Total Requests: 40,790
Network Traffic Ingress: 128MB
Network Traffic Egress: 815MB

[baljir0901]

Hey! 👋 I’ve run into this error before — "FATAL: the database system is in recovery mode" — and it usually indicates that PostgreSQL is momentarily restarting or recovering from an issue, such as:

• A crash or panic (even brief)
• High connection churn triggering forced restart
• Low-memory kill event (OOM)
• Disk-related hiccups

---

🧠 Key things to know

• Even if it only lasts 1–2 seconds, it’s a sign that Postgres went into recovery mode briefly and had to restart internal processes.
• The message "Hot standby mode is disabled" suggests it’s not a read replica issue — it’s happening on the primary.

---

📊 Your Supabase stats look low stress-wise

• CPU: ~3.5%
• Memory: ~56%
• Connections: Max ~30
• Disk IO: very low
• DB size: minimal

So this is likely not a performance pressure issue. But with a MICRO compute plan (1 GB RAM), it's possible your Postgres process is being OOM-killed (out of memory) when a burst happens.

---

🛠️ What you can try

1. Enable logs if possible
   Supabase logs don't show full PostgreSQL internal logs, but you can:

   • Check “Database Logs” tab for any OOM, PANIC, or FATAL lines.
   • Open a support ticket (if you're on a paid plan) to request deeper logs or insights.

2. Reduce connection bursts

   • Set a connection pooler like pgbouncer if you're not using Supabase's built-in one.
   • Use Supabase’s Connection Pooler (pgbouncer) and limit max connections in your app.

3. Consider upgrading to SMALL compute

   • More memory helps avoid OOM kills or brief panics.
   • Even 2 GB vs 1 GB can make a big difference in stability during bursts.

---

Hope this helps clarify what's going on and gives you a few things to try — I know how frustrating intermittent DB restarts can be. Let me know if you want help setting up connection pooling or monitoring tools! 🙌

[sbutlerjr]

I upgraded the instance to SMALL yesterday as suggested. This problem persists and has happened 3 times again already today.