01 / CONNECTION LAYERS
How protocol, transport and TLS settings fit together
Start by identifying which layer each setting belongs to
Choosing a protocol in Add Server determines how the client and server identify each other, specify the destination address and exchange data. It doesn’t determine which websites you visit or choose your routing strategy. A request usually also involves local DNS, routing rules, the underlying transport and the remote exit. Separating these layers helps explain why a connection can work with one protocol but fail during the handshake because a domain field is wrong.
Server is the remote address, and Port is the port it listens on. The server address can be a domain or an IP address, but they aren’t always interchangeable. A domain may be used for address routing, while TLS verification may require a specific hostname. Replacing a domain with an IP can still reach the machine, but that doesn’t mean the certificate, HTTP routing or backend service will match. Keep the original details so you have a basis for comparison.
Authentication, transport and certificate verification serve different purposes
Authentication settings prove that a connection is allowed, commonly using a password, UUID or public key. Transport determines how protocol data is carried, such as directly over TCP or inside WebSocket. TLS establishes an encrypted channel and verifies the server’s identity. Some protocols provide encryption themselves, while others rely mainly on an outer secure channel. A password and TLS are not the same thing, and one cannot be inferred from the other’s name.
With HTTP-based transports such as WebSocket, Host and Path often help route requests on the server. SNI is sent during the TLS handshake to indicate the intended hostname. These values may be the same or may differ depending on the deployment. Changing all three to Server may look consistent, but can send requests to the wrong virtual host or path. If your subscription provides these settings, keep them as given unless you receive updated configuration details.
| Layer | Common settings | What a mismatch can cause |
|---|---|---|
| Connection endpoint | Server、Port | Name resolution failure, connection refused or timeout |
| Protocol authentication | Password、UUID、Key | Authentication failure or connection closed |
| Transport and routing | Transport、Host、Path | Connection closed after handshake, or incorrect HTTP routing |
| Secure channel | TLS, SNI and certificate verification | Certificate name mismatch or trust chain failure |
A successful import only means the data could be read
Seeing server entries after importing with Subscribe only means the client could read the subscription data. It doesn’t prove the remote port is reachable or the credentials are valid. Sharing formats also differ in which settings they can represent: a short link may contain authentication details but omit additional transport settings. If a manually added entry connects but the subscription entry doesn’t, compare settings layer by layer instead of repeatedly switching protocols. The selected protocol must match the server implementation and listener configuration.
TCP and UDP also need to be understood at separate layers. An app using UDP doesn’t necessarily mean the client-to-server transport must also use UDP; some proxies encapsulate datagrams. Conversely, a UDP-based tunnel can carry TCP traffic. Check the needs of the destination app, the client’s protocol capabilities and the server’s forwarding settings. Don’t assume the entire route works just because “UDP is enabled.”
When troubleshooting, write a brief description of the connection: which endpoint, which protocol, how it authenticates, what transport it uses and whether TLS verification is enabled. This is more useful than noting a server name, which can be set to anything in a subscription and may not reflect the actual settings. The sections below explain these components by protocol. If an advanced option isn’t available in the app, don’t substitute a similar-looking setting; check the options actually provided by your client and the original configuration details.
02 / SHADOWSOCKS
Shadowsocks: Match the encryption method and password
Understand its trade-offs as a streamlined proxy-forwarding model
Shadowsocks was designed as a relatively lightweight encrypted proxy between a client and server. The client sends the destination address and traffic to the server, which then connects to the destination. Unlike a full network-layer tunnel, it handles proxy sessions rather than assigning the device a complete virtual network identity. This keeps the basic configuration focused, but not every connection called Shadowsocks necessarily uses identical authentication or data formats.
The encryption method is the first setting to check. Earlier stream ciphers and later AEAD methods differ in message authentication, key use and packet structure. They can’t connect to each other just because the password is the same. Use the method specified by your existing server configuration; don’t assume that a higher bit count means a method will be faster or better on a phone. Performance depends on processor acceleration, implementation efficiency and data volume. The name alone can’t provide a reliable speed ranking.
Password format is more than whether the text is readable
In traditional configurations, the password is typically used to derive a key. Some newer methods require a specific key length and encoding. A value that looks like a random string may be an encoded key of a fixed length; don’t shorten it, add spaces or replace it with an easy-to-remember word. When copying, watch for a trailing newline, full-width punctuation or spaces inserted by a messaging app. If the client reports a format error, check the original configuration instead of cycling through encryption methods.
When adding a server manually, first check Server, Port, the encryption method and Password. Then confirm whether any additional transport settings are actually required. Some deployments add processing beyond the base protocol, so the parsed sharing link must match the server. Don’t add TLS, Host or Path to an ordinary Shadowsocks configuration just because another guide mentions them. Adding an encapsulation the endpoint doesn’t use makes the two sides expect different data and can cause the connection to close immediately after it opens.
Test web traffic, datagrams and persistent connections separately
Test Shadowsocks TCP traffic and UDP forwarding separately. A webpage loading only shows that the request path used by that page received a response. Voice calls, games or some DNS requests may still be affected by server UDP settings, network-device timeouts or destination port restrictions. First check the transport used by the failing app, then confirm that the remote server allows the corresponding forwarding. A webpage speed test is not a substitute for checking datagram traffic.
On iPhone and iPad, a relatively lightweight encapsulation can reduce the processing cost of forwarding each request, but battery use still depends mainly on wireless conditions, the number of connections and retransmissions. A distant server or high packet loss can keep the wireless radio active longer, even if the protocol itself requires few processing steps. Conversely, on a stable network, an extra security layer may make no noticeable difference to battery life. Compare under the same conditions rather than relying on impressions.
If a test occasionally succeeds but downloads stall, note whether the interruptions occur only during heavy traffic, after an idle period or after switching Wi-Fi. Heavy-traffic issues may involve congestion or server resources; delays after inactivity may mean an intermediary has closed the connection; failures after switching networks may mean the old session hasn’t been re-established. None of these symptoms directly proves the password is wrong. Authentication mismatches usually recur consistently, so compare the stage shown in logs with results from repeated tests.
For an existing working setup, choose Shadowsocks because the server supports it, its settings are clear and it passes tests with your actual apps—not because it’s automatically better than other protocols. Keep the original entry and create a separate test entry, changing only one related group of settings at a time. Replace the default server only after checking that webpages, required UDP traffic and recovery after idle periods all work. This preserves a way back and helps show what made the difference.
03 / VMESS & VLESS
VMess and VLESS: Keep identity, encryption and transport distinct
The two protocols divide responsibilities differently
VMess incorporates user identification, protocol authentication and data protection into its design. It commonly uses a UUID as the user ID and can work with different underlying transports. VLESS uses a leaner division of responsibilities and doesn’t provide the same built-in data encryption layer as VMess; deployments typically rely on an external security mechanism such as TLS. Both may use UUID, TCP or WebSocket settings, but similar-looking fields don’t make the protocols interchangeable. The server must explicitly support the selected protocol.
A UUID identifies a user; it isn’t a server domain or a device identifier, and you shouldn’t replace it with an arbitrary new value. The client and server must have matching IDs. If the provider changes the user ID, an old entry may fail authentication even when its port is still reachable. After updating with Subscribe, make sure you’ve selected the updated entry. Similar old and new names can make it look as if the settings have changed when they haven’t.
Understand legacy settings; don’t reuse outdated values blindly
Older VMess configurations may include alterId. It belongs to an earlier authentication design; it isn’t a concurrency setting you can increase to improve speed. Newer AEAD configurations typically use a value of zero. Whether this setting appears and what it means depend on the implementations at both ends. For an existing configuration, follow the server’s instructions rather than filling in a number found online. Keep the device clock accurate as well, since authentication may involve time checks and clock drift can complicate diagnosis.
An encryption setting in a VLESS configuration and the outer security channel are separate concepts. A field set to none doesn’t mean you can also disable TLS if the configuration requires it. Check the protocol setting, TLS toggle, verification name and transport separately. If the server requires an additional Flow or a specific security setting, confirm that the client supports it. If there’s no matching option, don’t try to substitute a note, Path or other input field.
| Setting | VMess | VLESS |
|---|---|---|
| User identity | Usually a UUID | Usually a UUID |
| Data protection | Built into the protocol | Usually provided by an outer secure channel |
| Legacy compatibility setting | May include alterId | Don’t reuse legacy VMess settings |
| Checks common to both | Endpoint, transport, TLS, SNI, Host and Path must match the server | |
WebSocket and TLS are additional compatibility requirements
With WebSocket transport, a connection typically reaches the server or an intermediary first, then Host and Path route it to the appropriate backend. Path capitalization, the leading slash and any parameters may matter; don’t paste an entire sharing link into the Path field. Set SNI and the certificate verification name according to the configuration as well. If the certificate is valid but the connection then closes, check transport routing and authentication instead of assuming every TLS-related setting is correct.
Direct TCP transport and additional HTTP encapsulation have different handshake and processing costs. Actual page speed also depends on connection reuse, server deployment and network round-trip time. A slow first visit followed by faster ones may simply reflect DNS, handshakes and caching. When comparing VMess and VLESS, changing the server region, port and transport at the same time makes it impossible to attribute the result to the protocol. Keep other conditions as consistent as possible.
On iOS, UUID authentication doesn’t inherently use more battery than password authentication, and VLESS doing less protocol processing doesn’t guarantee longer battery life. Look for repeated reconnections, excessive testing or background activity that could amplify the difference. For help identifying settings, read VMess vs. VLESS. When adding a server, use the complete settings from your existing configuration rather than treating example values in an article as a working setup.
04 / TLS & APPLICATION PROXY
Trojan, HTTP and SOCKS5: Different security boundaries
TLS is a key part of Trojan
A typical Trojan deployment runs the proxy protocol inside a TLS channel and uses a password for access authentication. Check two things: whether you reached a server with the correct certificate, and whether the server accepts the password. A successful TCP connection only proves that the port responded; completing TLS only proves that the secure-channel stage passed. Protocol authentication or backend forwarding can still fail. Breaking the process into stages makes troubleshooting more precise than saying simply that “the server is down.”
Server, Port, Password and the TLS hostname should come from the same configuration. The IP used to connect and the certificate’s domain may differ, but you must set the verification name as required by the deployment. Handle an expired certificate, incorrect device time, incomplete trust chain or hostname mismatch as separate issues. Skipping certificate verification weakens server identity checks and shouldn’t be a routine fix. First check the clock, domain and original configuration; then ask the server maintainer to address certificate issues.
An HTTP proxy and an HTTPS website are different things
HTTP proxies were originally designed to forward application-layer requests and are now widely used with the CONNECT method to create a byte tunnel to a destination. Visiting an HTTPS website doesn’t mean the connection between your client and the proxy server also uses TLS. With a standard HTTP proxy, authentication details and connection metadata on that leg may still be visible to someone monitoring the network. Only a proxy connection explicitly protected by TLS extends the secure channel to the proxy endpoint.
HTTP proxies commonly authenticate with a username and password. A server may also restrict which destination ports can be used with CONNECT. If some webpages work but an app fails completely, check whether its traffic is outside the proxy’s capabilities or access rules before changing credentials. On a managed network that provides only a specific proxy type, use the type supplied by its administrator. A familiar-looking port doesn’t make an HTTP proxy a Trojan or another encrypted protocol.
SOCKS5 negotiates proxy connections but doesn’t encrypt traffic
SOCKS5 provides a general way to negotiate proxy connections and specify destination addresses. It can also handle datagrams through the appropriate mechanism. A username and password control access, but don’t give the protocol content encryption. If the destination website uses HTTPS, its TLS protects the application content, but other information on the proxy link needs separate consideration. Use a standard SOCKS5 endpoint only when you understand the network boundaries and any additional protections in place.
Where a domain is resolved is easy to overlook with HTTP and SOCKS5. The client may resolve the address first or pass the domain to the proxy, depending on client settings, protocol capabilities and routing rules. These approaches can return different addresses and affect content delivery and access. If a domain fails but a test using its IP succeeds, check DNS and the verification name first; don’t assume proxy authentication has failed. An IP test may also bypass important conditions.
Choose among these three types based on the server endpoint and what your apps need. To troubleshoot Trojan, check TLS first, then the password. For HTTP, check the proxy connection type, authentication and CONNECT restrictions. For SOCKS5, check negotiation, domain handling and UDP support. A successful response from a port doesn’t replace these checks. If the server provides only one protocol type, changing the client’s selection can’t convert it to another.
On iPhone, brief web browsing may not reveal a noticeable battery difference between these options. Repeated TLS handshakes and retries after failed connections can increase wireless activity. If problems persist, restore the correct settings and reduce unproductive retries before comparing protocol processing costs. For sensitive traffic, also check that the destination app uses a secure connection. A proxy protocol doesn’t replace account verification, app permissions or the destination website’s certificate checks.
05 / NETWORK TUNNEL
WireGuard: Public keys, tunnel addresses and routing ranges
From proxy sessions to network-layer tunnels
WireGuard is built around network-layer tunneling. It identifies peers using public keys and exchanges encrypted packets over UDP. Its configuration model differs from a proxy entry with a server and password: as well as a remote endpoint, it needs a local private key, a peer public key, an internal tunnel address and an allowed address range. You can’t move a set of ordinary proxy subscription credentials directly into WireGuard, and a server public key alone isn’t enough to connect.
Private Key identifies you locally and should be kept secret. Public Key identifies the corresponding identity and can be shared with the party configuring the peer. The Peer Public Key you enter in the client usually refers to the server’s public key, not your own. The server must also register the client’s public key and address. If you generate a new private key without updating the remote server, the endpoint may be reachable but bidirectional communication won’t work. Changing DNS repeatedly won’t fix an identity mismatch.
The tunnel address isn’t the remote server address
Endpoint is the remote address and port reachable over the public internet or current network. Address is typically the local address used inside the tunnel. Putting the server’s public IP in the internal address field, or reusing an internal address that should be unique to each device, can cause routing and return-traffic problems. If an existing configuration includes both IPv4 and IPv6, preserve them as designed and check that both ends support them. Don’t delete IPv6 tunnel settings just because your local network shows IPv4.
In WireGuard, AllowedIPs helps determine which addresses belong to a peer and which peer handles outgoing traffic. In Shadowrocket, it must also be understood in the context of the client’s own rules and connection mode. It isn’t a “speed-up list” that can be expanded without context. A range that’s too narrow may omit destinations you need; one that’s too broad may route addresses through the tunnel that should connect directly. First find out what the remote end is allowed to carry, then decide how the client should route traffic.
| Setting | Meaning | What to check |
|---|---|---|
| Private Key | Private key for the local identity | Must match the client public key registered on the remote end |
| Peer Public Key | Peer identity | Don’t enter your own public key |
| Endpoint | Address and port used for the connection | The underlying network must be able to reach the UDP endpoint |
| Address / AllowedIPs | Tunnel address and address range | Must match server routing and client traffic rules |
Adjust keepalive and MTU to address a specific issue
Behind a NAT device, an idle UDP mapping may be reclaimed. An appropriate Persistent Keepalive can help maintain it. More frequent keepalives aren’t always better; excessive packets may make it harder for a mobile network to enter a low-power state. Follow the deployment requirements. If only the first request after a long idle period is slow, confirm that the mapping is the cause before enabling keepalives. They aren’t a universal fix for connection failures.
MTU sets the maximum packet size sent inside the tunnel. With additional encapsulation underneath, a value that’s too large can cause fragmentation, packet drops or path MTU discovery issues. A common symptom is that small requests work while large pages or uploads stall. A value that’s too small increases packet count and processing overhead, so don’t copy a supposed best value without testing. Save the original settings, then make gradual changes on the network in question. Treat MTU as a likely cause only if the issue changes consistently with your adjustments.
WireGuard can update endpoints when network addresses change, but after a phone switches between Wi-Fi and cellular, the system interface, address mapping and actual reachability may still take time to recover. It can’t guarantee uninterrupted service in every environment. Use it when you already have the correct peer settings, a clear need for tunnel routing and confirmed UDP reachability. If you need only a single proxy endpoint, a full network-layer configuration may add unnecessary maintenance and troubleshooting work.
06 / QUIC TRANSPORT
Hysteria2: UDP paths, congestion control and bandwidth settings
Designed for lossy links doesn’t mean faster on every link
Hysteria2 builds proxy transport on QUIC, relies on UDP underneath and makes design trade-offs for congestion control and data delivery over lossy links. Its behavior differs from traditional TCP transport, but that doesn’t mean it should be the first choice on every network. If the network restricts UDP, handles sustained UDP traffic differently or the server lacks resources, the protocol’s design can’t remove those external limitations.
QUIC provides reliable transport and TLS security within the protocol. It can reduce some forms of blocking between streams, but each reliable stream still has to recover lost data. With severe packet loss, retransmissions, congestion responses and competition for wireless capacity remain. Don’t confuse “different streams don’t have to wait for the same missing TCP byte” with “packet loss doesn’t affect traffic.” The latter isn’t true and shouldn’t guide your choice.
Match authentication, SNI and any additional settings
When adding a server, confirm that its type is Hysteria2, not a similarly named protocol or an older format. Server, Port, authentication details and TLS name must match the remote configuration. Some deployments also use additional data-handling settings; whether you need them and how to enter them depends on your existing configuration. Don’t enable an optional field just because it’s available. If the client sends data in a format the server doesn’t expect, the connection may fail even when the address and credentials are correct.
Certificate verification is still part of confirming server identity. If you connect by IP but the certificate is for a domain, set the correct name rather than disabling verification. Authentication and TLS failures occur at different stages: for authentication, check credentials and server permissions; for TLS, check the device clock, certificate validity and hostname. Focus on the underlying path only if the UDP port can’t be reached. Changing several settings at once makes even a temporary success hard to explain.
Bandwidth values are control inputs, not speed-test results
Some configurations have upload and download bandwidth settings that influence sending behavior. The values you enter don’t equal the actual speed available and can’t increase network or server capacity. Setting them far above what the path can handle may cause more queuing and packet loss; setting them too low may limit throughput. Check the units and whether upload or download is measured from the client’s perspective. Don’t confuse bytes per second with bits per second.
Mobile network capacity varies with signal strength, cell load and movement. A bandwidth setting that works on steady Wi-Fi may be too aggressive on cellular. If webpages become noticeably less responsive during uploads, the upstream queue may be saturated; the issue may not be download speed. Test lightweight requests during uploads and interactions during downloads to see which direction is affected, then adjust within the limits of your existing configuration.
On iOS, sustained high throughput keeps the wireless radio and processor working, so faster downloads and higher momentary power use can happen at the same time. To assess battery life, compare the total time and battery used to complete the same task, not just whether the device feels warm during use. For persistent background connections, check whether the client repeatedly reconnects while idle. If the link is unstable, additional traffic and handshakes may outweigh any processing-efficiency gains.
Consider Hysteria2 when you already have a compatible server, the current network allows the required UDP traffic and tests show a real benefit for your apps. If a stable TCP connection already meets your needs, there’s no reason to switch just because the protocol is newer. Keep the old entry before switching, then test webpages, persistent connections, uploads and recovery after network changes. Record exactly what improved rather than generalizing a result for one task to the whole connection.
07 / IOS MEASUREMENT
iOS speed and battery life: Compare equivalent tasks
Track time to first response, throughput and reliability separately
“Fast” can mean several different things: how quickly content starts arriving, the rate it sustains, how reliably it performs over time and how quickly it recovers after a network change. Switching protocols may improve one metric without affecting another. For web browsing, DNS and handshake delays matter; for large files, sustained throughput matters more; voice traffic depends on jitter and packet loss. Decide what you primarily use the connection for before choosing a meaningful test and observation period.
The latency test or Connectivity Test in Home is a diagnostic starting point, not a complete assessment of the entire route used by an app. The test target, connection method and current app traffic may differ, so a result may only show that one stage responded. After a short test succeeds, use the actual app for a while and verify that traffic follows the intended policy. Don’t treat one low latency reading as proof that a protocol is better, and avoid repeated tests that interrupt normal use.
Hold conditions steady and change one variable at a time
For a useful comparison, use the same device, network, approximate time and destination, while keeping the server location and load as similar as possible. If different protocols require different servers, acknowledge that the results include server differences rather than claiming to have measured the protocols alone. Repeat each task and compare first visits with visits after caching. Retest occasional anomalies; only consistent, reproducible differences should influence your final choice.
Control device conditions during tests. Screen brightness, background photo sync, app updates, charging and device temperature all affect battery use and throughput. Heat can change how the processor and wireless radio operate, meaning the later option may be tested under different conditions. You don’t need laboratory precision, but pause obvious background transfers, keep brightness similar and compare when the device is in a similar state. Otherwise, blaming a temperature difference on an encryption algorithm can lead to the wrong conclusion.
Wireless activity often matters more than protocol processing
Phone battery use isn’t just about encryption and decryption. Weak signals, repeated retransmissions, frequent keepalives and repeated wakeups can all prevent the wireless radio from entering a low-power state. A protocol with fewer processing steps may use more power than a more complex one if it continually rebuilds the connection. Observe both idle and active periods, distinguishing data transfer, background activity and connection maintenance. Don’t judge by battery use during a single download.
On Demand starts a connection when specified conditions are met; it doesn’t increase protocol throughput. If its conditions repeatedly change with network state, the connection may start and stop too often. For troubleshooting, establish a baseline using a clear manual connection state, then restore automatic conditions. Similarly, routing more traffic through a proxy changes the amount of data forwarded. Keep Global Routing mode and the selected Config consistent when comparing battery use.
| Use case | What to measure | Common confounding factors |
|---|---|---|
| Web browsing and lightweight apps | First response and repeat visits | DNS, cache and destination website status |
| Sustained file transfers | Consistent throughput and time to complete | Server load, temperature and queue congestion |
| Voice and interactive apps | Jitter, interruptions and recovery | UDP support, wireless handoffs and packet loss |
| Idle background use | Reconnect frequency and keepalive activity | Background sync and On Demand conditions |
When switching between Wi-Fi and cellular, the IP address, DNS, default route and intermediary address mappings may all change. Verify that each network works on its own before testing the transition. Otherwise, you can’t tell whether one network was already unavailable or the session failed to recover. If the problem occurs only after switching, disconnect and reconnect to see whether it recovers, and record the network conditions before and after. Restarting the app may temporarily help, but can obscure useful clues about the failure stage.
Keep your test notes simple: protocol and transport, network, task, result and conditions needed to reproduce the issue. Don’t share screenshots or logs containing complete authentication details. For DNS issues, see DNS Settings Explained. For the first test, see First Connection and Connectivity Test. Recording results for a specific task is more useful than a vague note saying “faster” or “uses less battery.”
08 / SELECTION & VERIFICATION
Choosing a protocol, checking rules and troubleshooting by layer
Check compatibility first, then compare the experience
Choose in this order: what the server supports, what the current network allows, what your apps need, and only then your performance preferences. If your existing subscription provides one protocol, the client can’t change it into another. If you have multiple valid endpoints, first rule out incomplete settings, capability mismatches and unreachable underlying connections. Then compare them using the method in the previous section. Neither the protocol’s age nor the number of settings is a useful first-pass selection criterion.
For everyday browsing and apps, start with an endpoint that’s already proven stable and has clear settings. For UDP traffic, confirm that the protocol, server and current network all support it. For access to a specific private network, the tunnel’s routing may matter more. If the only available endpoint uses HTTP or SOCKS5, first understand its security boundaries. This is a decision-making method, not a fixed ranking; the same protocol can behave very differently on different networks.
After the protocol connects, check Global Routing
The three Global Routing modes—Config, Proxy and Direct—work differently. Config applies the rules in the current configuration; Proxy sends traffic handled by the client through the proxy in that mode; Direct connects directly. The scope still depends on system-level routing and app behavior. Temporarily switching to Proxy can help isolate a rules issue, but don’t leave the diagnostic mode in an unknown state or later tests won’t have the same baseline.
This short Shadowrocket rule example illustrates the order of domain, geographic IP and final policies; it isn’t a complete server configuration. It assumes the PROXY policy in the relevant configuration points to a working server. Rules are evaluated from top to bottom and generally stop at the first match. FINAL handles traffic that matches none of the earlier rules. A broad rule placed first can prevent later, more specific rules from taking effect; this is unrelated to whether the protocol itself works.
[Rule]
DOMAIN-SUFFIX,example.com,PROXY
GEOIP,CN,DIRECT
FINAL,PROXY
Domain rules need domain information to match, while address-based rules may depend on DNS results. If only some websites fail, check the matched policy, DNS result and destination connection before changing the protocol on every server. GEOIP relies on the address database in use; it doesn’t guarantee that a website’s content, company location and exit route are the same. For limits on rejecting requests by domain, see How to Write REJECT Rules.
Troubleshoot the failing stage; don’t change every setting at once
First, check whether Server resolves and the port is reachable. Next, check the TLS name, certificate and device clock. Then verify the password, UUID or public key. Compare Transport, Host, Path and other transport settings after that. Finally, check destination resolution, routing and remote forwarding. Not every protocol has every stage, so skip what doesn’t apply. After each change, repeat the same test and record what changed. This avoids mistaking an occasional recovery for a successful settings fix.
If every entry fails, check the local network, system VPN status and current configuration first. If only one entry fails, focus on its endpoint and credentials. If webpages work but voice traffic doesn’t, check the datagram path. If small requests succeed but large transfers stall, consider packet loss, congestion and MTU. If failures follow an idle period, look at connection expiry and keepalives. These branches help narrow the search; no single symptom proves the cause. Support each conclusion with another test.
Back up entries managed by Subscribe before making changes, since a later update may overwrite local edits. Create a separate copy for testing, and keep credentials and logs somewhere secure. When reporting an issue, include only the protocol type, failure stage and redacted settings structure. If you’re unsure about the basics, return to the Quick Start guide. For download or installation issues, see the installation instructions in the Download Guide. Finish with a result you can reproduce, not a string of protocol changes.