When Shadowrocket is slow, first distinguish a long initial load from consistently slow downloads or intermittent disconnects, then test one fixed target at a time. This guide assumes you already have your own provider, nodes, and subscription. It focuses on the Home, Config, and Settings screens on iPhone and iPad; the same troubleshooting sequence applies to Mac, Windows, Android, and Linux, though control locations may vary. The goal is to document reproducible conditions—not to keep changing settings based on a single latency test.
1. Set a baseline: low latency doesn’t mean fast downloads
Connectivity Test on Home measures connectivity and response time using the selected test method; it does not measure sustained throughput. A connection can come up quickly while a congested node exit still causes video buffering. If the first DNS lookup is slow but subsequent downloads are fine, check DNS first. Start by noting exactly when the slowdown occurs; that’s more useful than switching protocols straight away.
- Choose one fixed webpage you’re authorized to access and one fixed download. Keep both targets unchanged during testing, and don’t run other speed tests at the same time.
- Record whether you’re on Wi-Fi or Cellular, which node is selected on Home, the Global Routing mode, the test time, and whether other network-filtering tools are enabled.
- Repeat the test three times under the same conditions, observing sustained transfers for about 30 seconds each time. Record the time to first response, the speed after it stabilizes, and any interruptions. This is a suggested test method, not a claim about measured device performance.
- Change one variable per round. When switching nodes, keep the local network unchanged; when switching networks, keep the node unchanged. Restore your usual settings when you’re done.
Keep Mbps and MB/s distinct: 8 Mbps is about 1 MB/s before protocol overhead. Browsers, download tools, and network plans may use different units, so don’t compare the numbers directly. Download-source throttling, device storage write speeds, and a video platform’s automatic quality selection can also affect what you see.
2. Check the traffic path first: Global Routing and rule matches
On Home, open Global Routing and note the current setting before changing it. Config routes connections according to the active configuration; Proxy sends traffic through the selected proxy policy or node; Direct connects without a proxy; Scene routes traffic according to the configured scene. With Scene, check which policy the active scene actually applies. A selected node alone doesn’t mean the request is using it.
For a destination that normally allows direct connections, briefly compare Direct and Proxy. If Direct works but Proxy is slow, investigate the proxy path. If both are slow, consider both the local network and the destination service. A failed Direct connection to a destination that requires a proxy does not prove the local network is faulty. When testing is complete, restore Config or your previous Scene.
DOMAIN-SUFFIX,example.com,PROXY
IP-CIDR,192.168.0.0/16,DIRECT,no-resolve
GEOIP,CN,DIRECT
FINAL,PROXY
These examples are for reading rule syntax only; they are not a complete configuration to paste in. DOMAIN-SUFFIX matches domain suffixes, IP-CIDR matches address ranges, GEOIP uses IP geolocation data, and FINAL handles requests that matched no earlier rule. Rules are evaluated in order, and PROXY in the example must resolve to a valid policy in your active configuration. When checking the current configuration in Config, look for an earlier rule sending the target through an unexpected route. Don’t delete all your rules just to run a speed test.
Takeaway: verify the exit route before judging node speed
If the same destination works in Proxy but is slow in Config, check which rule and policy the request matches next—not the node’s protocol.
3. Node checks: connection failures, congested exits, and subscription changes
Start by checking what’s selected, whether its settings are valid, and whether the service is available. Confirm the selected node on Home, then open its edit screen and check Address, Port, protocol, and authentication fields. Use the port provided by your own service provider. HTTPS commonly uses port 443, but that does not mean you should change an arbitrary node’s Port to 443.
- Connection fails immediately: Check whether Address resolves, whether Port matches the server, and whether the authentication details have expired. If a node in your subscription has been replaced, update the contents of your existing subscription.
- Pages load, but transfers stay slow: Keep the target fixed and compare it with another node in your existing subscription. If only one node remains slow, note the node and time of day; don’t change global DNS settings first.
- Only one site is slow: Test another destination. The route from the node’s exit to the site, destination-side throttling, or regional policies may affect just one site.
- The problem started after updating a subscription: Recheck the node selected on Home and the policy-group references. A successful update only means the subscription contents were retrieved; it doesn’t mean every node in it is reachable.
A subscription and a Config profile are different layers: the former typically provides node details, while the latter can include rules, policies, and DNS settings. Don’t update both while troubleshooting, or you won’t know which change affected the result. Note your current selections before updating, and keep a recoverable copy before editing a configuration.
Error: connection refused
Cause and fix: This usually means the destination host or a device along the path actively refused the connection. Check the node’s Address and Port, then ask your own provider to confirm that the service is available on that port. This is not simply a bandwidth issue.
Error: i/o timeout
Cause and fix: A read or write did not complete within the allowed time. This message alone can’t tell you whether DNS, the node, or the destination is at fault. Check the address and operation stage in nearby log entries, then compare results by changing the node and the network separately.
Log wording can vary by operating system and connection stage. Keep the original error and surrounding context; that’s more useful than paraphrasing it as “timeout.” Before sharing logs, redact tokens in subscription URLs, authentication fields, and any other sensitive information.
4. Route checks: compare networks and times to identify congestion
Here, the route includes two segments: from your local carrier to the node, and from the node’s exit to the destination site. A region in a node’s name says nothing about the quality of the entire route, and a shorter geographic distance doesn’t guarantee more bandwidth at peak times. To isolate route issues, compare performance across different times and network connections.
| Test condition | What you may observe | Next step |
|---|---|---|
| Same node and destination; switch between Wi-Fi and Cellular | Only one network connection is slow | Check that network or its route to the node; don’t change authentication fields yet |
| Same network and destination; switch to another existing node | Only one node remains slow | Check that node’s load, exit, and route to the destination |
| Same network and node; test at different times | Performance repeatedly drops during busy periods | Record results from multiple time periods and report the congestion pattern to your own provider |
| Same network and node; try another destination | Only one destination is slow | Check the destination’s status, the rule-selected route, and any destination-side throttling |
If you use a route-diagnostic tool, packet loss at an intermediate hop does not mean that forwarded traffic is experiencing the same loss. Some routers limit diagnostic responses. Judge the result using the final destination’s response, actual disconnects, and sustained transfers; don’t request a route change based on one intermediate hop’s figures.
Takeaway: keep the node fixed and switch networks to narrow down the problem
If the same node works on Cellular but is slow on your home Wi-Fi, check the home connection and its upstream route first. Don’t rebuild the node, replace Config, and rewrite DNS settings all at once.
5. Protocol checks: TCP, UDP, and handshakes depend on server support
Shadowsocks, VMess, VLESS, Trojan, Hysteria2, and WireGuard use different transport and authentication methods. Changing Type in the client won’t make the original server support another protocol. Compare protocols only if your own provider offers the corresponding endpoint and complete settings. Speed tests across different exits also can’t isolate the protocol as the cause.
- Shadowsocks: Check that the encryption method and password match the server. UDP availability also depends on the server and the network path; a webpage loading successfully doesn’t guarantee that UDP traffic will work.
- VMess / VLESS: Check the authentication ID and transport settings against your existing configuration. If using TLS, also check fields such as the server name. Don’t disable certificate verification to mask a handshake failure.
- Trojan: Check the password and TLS-related fields. If the connection fails during setup, check the device time and server certificate instead of repeatedly increasing the number of speed tests.
- Hysteria2: It uses QUIC and requires a working UDP path. If your current Wi-Fi restricts UDP, compare results on Cellular first; don’t assume every failure is caused by an incorrect password.
- WireGuard: Check the keys, peer address, and related route settings. If small requests work but larger transfers stall, investigate MTU against your existing configuration, one step at a time.
MTU is not a universal speed boost. Consider changing it only when problems involving larger packets, nested tunnels, or a specific network can be reproduced consistently, and follow the server’s guidance. Record the original value and each result. Don’t apply one unverified fixed value to every device or assume a protocol guarantees a particular speed increase.
6. Local network and DNS: distinguish slow lookups from slow transfers
A long wait on the first visit followed by a much faster refresh may involve DNS caching, connection reuse, or TLS session resumption; that alone doesn’t prove a DNS problem. If sustained downloads are also slow, check wireless signal strength, upstream usage, and the access route instead of just cycling through DNS addresses.
- Rule out Wi-Fi issues: Keep the node unchanged, test closer to the router, then compare Wi-Fi with Cellular. Check whether a device is uploading a backup; a saturated upstream connection can slow download acknowledgements and webpage interactions.
- Check for competing network controls: Pause other VPN or network-filtering tools during troubleshooting to avoid overlapping traffic interception. Note their original settings so you can restore them afterward.
- Check where DNS comes from: Open the active configuration in Config and review its DNS settings. Distinguish system DNS, DNS configured in the profile, and DNS over the proxy path. Change one source at a time; don’t add several unfamiliar addresses at once.
- Test the domain correctly: Visit the same domain repeatedly and check the relevant logs. Don’t diagnose a DNS failure just because a direct request to an HTTPS IP address fails; the certificate, SNI, or virtual host may require the domain name.
- Rule out automatic switching: In Settings → On Demand, note the existing conditions. During a controlled test, temporarily disable automatic connection rules and keep the connection active manually from Home. Restore the original settings afterward.
Traditional DNS commonly uses UDP or TCP port 53, while DoH usually uses HTTPS; networks may handle them differently. Before changing the resolution method, confirm that the active configuration actually uses the setting. If the logs show the destination address resolving successfully but subsequent connections keep timing out, focus your checks on the node and route.
7. Common questions and a useful troubleshooting record
Latency is only a few dozen milliseconds. Why does video keep buffering?
A quick request doesn’t measure sustained bandwidth. Keep the node and network unchanged, compare sustained transfers to the same destination, then try another video or download target. If only one platform has issues, check the route to that destination and the platform’s status instead of sorting nodes by latency alone.
Proxy is faster. Should I leave it on?
Use it first to confirm the difference, then return to Config and check the rule and policy matched by that destination. Choose a long-term mode based on your needs; don’t mistake a temporary switch during troubleshooting for a permanent configuration fix.
Will switching protocols make it faster right away?
First confirm that you already have an endpoint using the protocol. Without one, changing Type will only create a settings mismatch. If you do have one, test repeatedly on the same access network and under as similar exit conditions as possible, then compare the results.
It only slows down when I get home and connect to Wi-Fi. Should I reinstall?
Keep the node fixed and compare with Cellular first. Then check router upload usage and the network conditions in Settings → On Demand. Reinstalling changes multiple variables, doesn’t replace a network comparison, and may erase useful configuration details from the time of the problem.
If these checks don’t identify the cause, write up steps that reliably reproduce the problem. Report node connection or exit issues to your own provider, and specific access-network issues to your local network administrator. “It’s slow” alone doesn’t identify the layer at fault. This guide does not provide, sell, or recommend nodes or subscriptions; it covers how to use and troubleshoot your existing configuration.
- Device and operating system; Wi-Fi or Cellular; when the issue occurred and roughly how long it lasted.
- Global Routing mode, node protocol, and the policy actually matched by the destination request.
- Results when switching networks with the same node, switching nodes on the same network, and changing destinations with the same node.
- Time to first response versus sustained transfer speed; any interruptions; and the original error text with sensitive details redacted.
- Settings changed during this test and whether restoring them helped. Don’t share subscription tokens, passwords, private keys, or other authentication details.