If you already have a working configuration and want to check whether a request used the expected server, note the Data readings before and after, then check Global Routing and the matching rule. The server view shows where traffic exits; the app view shows which app it came from. These readings are two views of the same traffic, so don't add them together or treat them as a substitute for your carrier bill.
Two ways to view traffic in Data
Server and app views show the same traffic from different angles. When an app's request enters Shadowrocket's processing scope, its route depends on the current routing mode and policy. A single request can increase both an app row and a server row: the app row identifies the source app, while the server row shows which server handled the traffic.
By server
- Grouped by
- Server entries listed in Data.
- What to check
- The upload and download readings for the same entry before and after a test.
- Attribution limits
- Multiple apps can use the same server, so an increase in one row doesn't identify which app caused it.
Record readings before and after switching servers. Don't use the historical totals of two entries to determine the current route.
By app
- Grouped by
- Traffic records in Data attributed to an app.
- What to check
- The change in the target app's readings before and after one controlled action.
- Attribution limits
- One app can access multiple domains; a single row doesn't correspond to one domain or one rule.
If an app doesn't have its own row, that alone doesn't prove it has no network activity.
Compare uploads and downloads separately. Opening a webpage also sends requests, so mostly download traffic doesn't mean uploads should be zero. Conversely, a server row may show little change without indicating a problem: the request may match DIRECT, be served from cache, or transfer less data than the interface can display.
Use before-and-after readings to exclude historical totals
- Note the starting readings. Keep the network type and current configuration unchanged. In Data, record the upload and download readings and units for the target app and server, along with the start time. There's no need to reset the statistics; you can calculate the difference from the existing readings.
- Perform one easy-to-identify action. Pause any background downloads you control, then request a known-size test file in one app. To repeat the test, use a different uncached file and make sure the download completes. Don't refresh a subscription during the test, as that can generate extra requests.
- Subtract readings in the same units. When the action is complete, return to Data and wait until the readings stop changing noticeably. Then subtract the earlier reading from the later one in the same row. If you reset the statistics during the test, or changing units leaves too few significant digits, record a new starting point instead of comparing readings across the reset.
For example, these figures are illustrative, not measured: server downloads rise from 24.0 MB to 29.2 MB, while browser downloads rise from 8.0 MB to 13.0 MB. The increases are 5.2 MB and 5.0 MB, respectively; don't add them to get 10.2 MB. Similar increases can suggest that the same action went through that server. The 0.2 MB difference doesn't by itself show proxy protocol overhead; it could also reflect other requests, different measurement times, or rounding.
How to check: measure idle traffic first
On the same network, leave the device idle for 30 seconds and record the natural increase in the target rows. Then run the test for about the same amount of time. If the readings increase by roughly the same amount during the test as they did while idle, retry with a larger uncached file rather than using that run to judge routing.
Check routing results in Config mode
Open Home → Global Routing and confirm the current mode. To verify rules in a configuration file, use Config; results in Proxy or Direct mode don't prove that a configuration rule matched. If you're using Scene, first check the routing settings the scene actually applies. The scene name alone doesn't show that a particular rule matched.
[Rule]
DOMAIN-SUFFIX,example.com,DIRECT
FINAL,PROXY
This minimal example illustrates rule order; it isn't a complete configuration. If a request's destination domain matches the example.com suffix and no earlier rule matches, it uses DIRECT. Requests that don't match an earlier rule are sent to PROXY by FINAL. The example domain illustrates the syntax; it isn't a test file address.
- In
Config, confirm which configuration is active and keep a copy of the original before editing. Check whether any earlier rule matches the target, especially a broader rule. Put specific rules before fallback rules; a rule placed afterward can't override a fallback that has already matched. - Make one new request to a test domain you control. Check the hostname actually accessed; redirects, images, and downloads may use different domains. An increase in the app's row doesn't mean every domain involved in the page used the same policy.
- Compare the destination and policy shown in the request details, then check whether the server readings increased accordingly. If you can only see the Data summary, limit your conclusion to “traffic was observed on this server” or “no noticeable increase was observed.” Don't infer which specific rule matched. Restore the original mode and configuration after testing.
Why Data readings differ from carrier bills
To compare usage, align the time period and network scope first. Data reflects traffic recorded by the client; your carrier records cellular usage counted on its network. They may cover different connections, traffic directions, or time periods, so don't compare a day's increase in Data directly with a full month's plan usage.
- Network scope: Client traffic recorded over Wi-Fi shouldn't be compared with cellular usage for the same period. Note the start and end times of the cellular test, and rule out any network changes in between.
- Connection scope: System or app traffic outside the client's statistics may still use cellular data. Background syncing can affect the carrier's usage increase even when foreground apps are closed.
- Measurement level: File size isn't the same as total network traffic. Protocol encapsulation, control messages, and retransmissions can make the amount transferred differ from the file size. The difference alone can't identify a specific overhead.
- Units and billing updates: Check whether each side uses base 1000 or 1024, and whether the carrier's usage figures have updated. When the interface shows only a limited number of decimal places, compare overall changes rather than expecting the last digit to match.
Common questions about Data readings
A webpage loaded. Why didn't the selected server's traffic increase?
Try again with an uncached destination, then check the mode under Home → Global Routing. In Config mode, check whether the destination uses DIRECT or whether the configured policy actually selected another server. Compare the increase on the target server with changes on other server rows.
Why is a server still handling traffic when I haven't used any apps?
Pause any known background downloads and syncing, then track changes in the app and server rows over an idle period. If the connection turns back on after you manually disconnect, check the auto-connect settings under Settings → On Demand. Automatic reconnection and ongoing traffic from a particular app are separate things; verify them separately.
Does an increase in an app's row mean all of its traffic uses the proxy?
No. Choose one action, such as downloading a file, note the destination it accesses, and check the routing result for that destination. Sign-ins, images, and file downloads may connect to different hosts; apply your conclusion only to the requests you've checked.