Hysteria2 is worth testing on networks where UDP is reachable and latency or random packet loss is high, but it is not guaranteed to be faster everywhere. This guide is for users who already have connection details from their service provider. It explains how the transport works, then covers adding a profile, checking routing, and troubleshooting in Shadowrocket. The protocol overview applies to iPhone, iPad, Mac, Windows, Android, and Linux. The interface steps below use iPhone and iPad as the reference; check the corresponding controls on other platforms.
1. What Hysteria2, QUIC, and UDP Do
Hysteria2 is a proxy protocol, QUIC provides its underlying transport, and UDP carries the outer packets between your device and the server. These terms are not interchangeable. A browser's TCP connection can be forwarded through Hysteria2, so using UDP for the outer transport does not mean it can only carry UDP applications.
QUIC integrates a TLS 1.3 handshake and provides streams, multiplexing, and loss recovery. Data that requires reliable delivery still needs to be recovered after loss; forwarding UDP does not mean every datagram is guaranteed to arrive. Encryption, reliability, and congestion control address different issues.
Use the port specified in your provider's configuration. Even when the port number is 443, a network that allows 443/TCP may not allow 443/UDP. Being able to open a standard HTTPS page does not prove that the outer Hysteria2 connection is reachable.
2. Where the Speed Comes From: Congestion Control and Multiplexing
Packet loss can result from queue congestion or random factors such as wireless interference. Congestion control that treats loss as a key signal to slow down may reduce its sending rate too early on the latter kind of link. Hysteria2's Brutal congestion control can work toward a configured bandwidth target and compensate for packet loss. This is one reason it can maintain throughput on some high-latency links with random loss.
Hysteria2 also supports bandwidth auto-detection-related modes. The control method actually used depends on the client implementation, bandwidth settings, and server policy. Selecting Hysteria2 alone does not guarantee that the same algorithm is always in use. Higher upload and download bandwidth values are not automatically better: exceeding the available capacity can increase queueing and packet loss, slowing web responses and other devices on the network.
- High latency: When round-trip time is long, recovery after a drop in speed can take longer. Well-managed sending can help reduce throughput fluctuations, but it cannot eliminate latency caused by physical distance.
- Random packet loss: Compensation can help maintain throughput when bandwidth is still available. If loss is caused by an already-congested link, sending more data may make performance worse.
- Multiplexing: QUIC's independent streams can reduce transport-layer head-of-line blocking between streams. A stream with lost data may still have to wait for recovery, and all streams continue to share the connection's bandwidth.
Bottom line: Check the connection before tuning bandwidth
First confirm that the UDP handshake and sustained transfer work. Then use the sustained rate measured on the same network as a reference for bandwidth settings. If raising the value increases latency without improving throughput, lower it again instead of raising it further.
3. Networks Worth Testing—and What to Troubleshoot First
Choose a protocol based on results from the same network and server, tested at similar times. A single latency test reflects the response to a particular probe; it is not the same as page load time or sustained download throughput. No single percentage describes Hysteria2's performance gains across all networks.
| Network condition | What to check | Recommended action |
|---|---|---|
| High latency with occasional packet loss | UDP is reachable, with bandwidth available for sustained transfers | Test Hysteria2 and record throughput and interactive latency. |
| Low latency and little packet loss | The current connection is already close to the available bandwidth | Compare stability and battery use first; don't assume a speed increase. |
| Websites load, but the protocol handshake times out | A working TCP connection does not prove UDP is reachable | Check the UDP port, network restrictions, and server listener. |
| Transfer starts fast, then slows noticeably | The bandwidth target may be too high, or queues may be building up | Lower the bandwidth setting and see whether latency drops. |
| Public Wi-Fi sign-in is incomplete | Network access authentication has not been completed | Complete network sign-in before testing the proxy connection. |
Mobile network handoffs, weak signal, and server load can all affect results. For a fair comparison, keep the test destination the same and record whether the connection succeeds, how quickly the page responds initially, and how sustained transfers perform. Do not attribute every speed change after switching servers to the protocol, or treat one successful test as proof of long-term stability.
4. Add and Check Hysteria2 Settings in Shadowrocket
Get subscriptions and connection details from your service provider. If you already have a Subscribe link, import it first and check the resulting server profile to reduce manual entry errors. Hysteria2 and earlier Hysteria do not use fully interchangeable settings; changing only the Type and reusing every old field may not work.
-
Verify the configuration source
Check the protocol name, server address, port, authentication details, and TLS settings provided by your service provider. For a subscription, go to Home → “+” → Type → Subscribe, enter your existing subscription link, and save. After updating, review the imported profile.
-
Select the protocol
To add a profile manually, go to Home → “+” → Type → Hysteria2. On the edit screen, enter the server address and port. The port must be the server's UDP listening port, not a management page port or a port used by another service.
-
Check authentication
Check the Password and any authentication details specified by your provider, character by character, and make sure there are no extra spaces at either end. If an SNI value is provided, enter it as given. SNI supplies the server name for the TLS handshake and should not be replaced with an arbitrary domain.
-
Check optional settings
Enter obfuscation settings only when your service provider specifically requires them. Enter upload and download bandwidth using the units shown in each field; Mbps and MB/s are not interchangeable—1 MB/s equals 8 Mbps. Do not disable certificate verification to hide a certificate problem.
-
Save and verify
After saving, return to Home, select the server, check Global Routing, and turn on the connection. Approve the system prompt the first time you connect. If a subscription update overwrites manual changes, check the subscription source rather than repeatedly editing the same profile.
5. After Connecting, Check Routing and Rules
Selecting a server does not mean every request will use it. In Home → Global Routing, Proxy, Direct, Config, and Scene represent different routing modes. To troubleshoot a single server, you can temporarily select Proxy to reduce rule interference, then restore your previous configuration. Direct is not suitable for verifying proxy forwarding. When using Scene, also check which route the active scene selects.
In Config mode, rules in the active configuration determine how connections are handled. The examples below show rule syntax, not a complete configuration file; the server or policy referenced by PROXY must exist and be valid in your configuration. To check whether a request uses Hysteria2, inspect the rule it matches and the resulting policy—not just the selected server name.
[Rule]
DOMAIN-SUFFIX,example.com,PROXY
IP-CIDR,192.168.0.0/16,DIRECT,no-resolve
GEOIP,CN,DIRECT
FINAL,PROXY
DOMAIN-SUFFIXmatches domain suffixes; in this example,example.comand its subdomains use PROXY.IP-CIDRmatches destination IP ranges. This example routes the specified private address range through DIRECT;no-resolvemeans the rule does not proactively resolve domain names.GEOIPmatches IP geolocation data. Results depend on the database records and cannot be inferred from the domain suffix alone.- Place
FINALat the end to handle connections that do not match earlier rules. Rules are checked in order; once a connection matches an earlier rule, the final policy is not used.
If the connection comes back after you disconnect manually, go to Settings → On Demand and check the automatic connection conditions. On Demand controls when a connection starts; it does not increase Hysteria2 throughput or fix incorrect authentication, certificates, or UDP ports.
6. Common Issues: Timeouts, Certificates, and Slow Transfers
Why do latency tests work when websites won't load?
The test and the request may take different paths. First check the selected server and Global Routing in Home, temporarily use Proxy to verify the route, and then review the error. If DNS resolution fails, check DNS. For a TLS handshake or authentication failure, review the relevant profile settings. A latency result alone does not prove that the protocol is working correctly.
Why does it time out on Wi-Fi but connect on mobile data?
Keep the same server and configuration while testing both networks. If only one Wi-Fi network fails, first check its sign-in requirements and UDP restrictions. If both fail, check the address, UDP port, and server status. A successful TCP port probe does not confirm UDP reachability.
Can I ignore a certificate error?
First check the device's date, time, and time zone. Then confirm that SNI matches the certificate name provided by your service provider. If the address is an IP, pay particular attention to the TLS name setting. Keep certificate verification enabled and correct the configuration. Ask your service provider to investigate any certificate issue you cannot verify.
Why do other pages slow down when a download uses all the bandwidth?
Pause high-bandwidth tasks first and see whether response times recover, then lower the bandwidth target in the profile. If latency rises sharply as a transfer starts, queues may be building up locally or upstream. Setting a higher bandwidth value does not increase link capacity and may affect other devices on the same network.
After troubleshooting, record the network type, the stage where the error occurred, whether switching networks resolved it, and any fields you changed. Do not post a full subscription link that contains authentication details in a public report. Whether Hysteria2 suits your network should be judged by repeatable connection success, interactive responsiveness, and sustained transfer performance together.