Why was my connection reset (TCP RST)?
A reset (RST) ends a TCP connection immediately, without the normal FIN close. When it arrives in the middle of a working connection, the app or a device in between gave up on it: a crash, a timeout, or a firewall or load balancer dropping idle connections. The packet shows who sent it, not why.
What it looks like in the packets
- An established connection (it had a SYN/ACK and data) ends with RST instead of FIN.
- The sender of the RST is the side (or the middlebox) that gave up.
- The other side reports "connection reset by peer".
Common causes
- The server app crashed, restarted, or hit a request limit.
- An idle timeout on a firewall, NAT, or load balancer.
- The app closes connections with resets by design (some do).
- A security device blocked the content.
What to check
- Look at which address sent the RST and how long the connection had been idle.
- Match the time against server and load-balancer logs.
- Compare idle timeouts on firewalls and load balancers with the app's keep-alive settings.
Wireshark display filters
tcp.flags.reset == 1
How Blame Finder flags it
Low for three or more reset connections; otherwise Informational. Upload a capture and the finding links straight to the packets behind it, with the most likely culprit at the top.