Node HandbookNODE HANDBOOK中文
v2rayN · Windows / macOS / Linux

v2rayN "Failed to run Core" in the log

The dialog is a summary of a failed start, not a diagnosis. The line that decides what to do next is the first error in the core log above it — and a port conflict looks nothing like a rejected configuration field. This page routes on that line instead of on the dialog text.

Node Handbook editorial team ·

Material checked on 2026-11-21. Error strings below are quoted verbatim so you can match them against your own log. This page routes on the first core error; it does not assume your nodes are broken.

The dialog is a summary, not a cause

Failed to run Core is what v2rayN shows when the core process it started did not survive. It is produced by the same dialog for a port that is still bound by the previous process, a generated field the selected core rejects, a missing data file, or a permission problem on a TUN device. The dialog text is identical in all of those cases; only the first line of the core log differs.

Two consequences follow. First, no fix can be chosen from the dialog alone — anything you change before reading the log is a guess, and you will not know afterwards which change mattered. Second, if the core never started, every server in your subscription will look unusable, because nothing is listening on the local port. That is a start-stage failure, not evidence about the airport.

Match the first core error to the failing stage

Log or symptomStage it points toDo this first
bind: address already in use, often with failed to listen TCPThe local listen port is still held by an earlier core or another clientSave the port number, quit the other instance normally, start one core only
A FATAL line naming a configuration field, JSON or YAML rejectionThe generated configuration and the selected core version disagreeRecord both versions and the full first FATAL line
Fails only when TUN is enabled; system proxy mode startsTUN device, routing or elevated permissionsTurn TUN off and confirm system proxy mode starts as a baseline
Core starts and listens, but pages do not loadProxy takeover, rules, DNS or the node itselfMove to the connected-but-no-internet checklist
No nodes at all after a subscription update, core never startsFetch, parse or import of the subscriptionMove to the empty node list checklist
failed to open geosite.dat or ... > EOFA Geo data file is missing or truncatedMove to the Geo file page and match the exact path in your log

If your first error is not in this table, do not start working down the list anyway. Each row assumes a specific piece of evidence; applying a fix without that evidence changes several variables at once and makes the next log unreadable.

Record these five things first

  1. The v2rayN version and the operating system version. "Latest" is not a version. The dialog and the core layout both change between releases.
  2. The selected core and its version — for example Xray or sing-box. The same configuration text is accepted by one and rejected by the other.
  3. What you did immediately before the failure: started the app, switched servers, enabled TUN, updated the core, or closed TUN.
  4. The first failure line in the v2rayN log. Later lines usually describe how the failure propagated.
  5. The first ERROR, FATAL or Failed to start line in the core output, quoted exactly.

Before you post a screenshot or a log excerpt anywhere, cover the subscription URL, the token, server addresses, UUIDs, passwords, private keys and your local user name. A port number and an error string are enough for someone else to follow the branch; credentials are not needed and cannot be unshared.

Port still bound: rebuild a single-core baseline

A public record on the v2rayN tracker covers one macOS case: after TUN was turned off and Xray restarted, the log showed failed to listen TCP on 127.0.0.1:10809 together with address already in use. That record supports a port conflict in that specific case. It does not prove that every "Failed to run Core" is a port problem, and it does not identify which process holds the port.

  1. Turn off TUN and turn off the system proxy.
  2. Quit v2rayN normally, then confirm a second v2rayN window or a second core is not running.
  3. Do not run two proxy clients that listen on the same local port at the same time.
  4. Reopen v2rayN and start a single core.
  5. If the same port is still refused, write down the port and identify the process holding it. Do not terminate system processes you do not recognise; on a managed device, hand it to the administrator.

Changing the local port only tests whether the conflict moves with the port. It neither identifies the holder nor proves that a stale process was the cause.

Configuration rejected: compare the version pair

One closed public report documents a specific pair — v2rayN 7.24.9 with sing-box 1.14.0 — where the generated DNS configuration failed sing-box validation and the core exited with a FATAL line. The maintainer's response pointed at a later v2rayN release. That is a record of one version combination, not a general rule, and this site has not verified the outcome of upgrading on other systems.

  1. Write down the v2rayN version and the core version as a pair.
  2. Check the official release notes for that pair before changing anything.
  3. Back up the configuration and any custom overrides.
  4. Update, then test with a minimal configuration before restoring your overrides.
  5. Restore one custom DNS, routing, override or TUN setting at a time.

The working configuration is often regenerated on every start. Editing a temporary file directly may hold only until the next launch and can introduce a JSON error of its own, so it is not a first choice for a normal user.

Fails only with TUN

Separate record, separate stage: one Windows report describes a system-proxy mode that still worked while only the TUN start failed after a sing-box update. That is enough to say that "system proxy works, TUN fails" should be judged independently of the nodes and the subscription. It is not proof that today's releases carry the same defect.

Turn TUN off and keep the same subscription, the same server and the same core. If system proxy mode starts and passes traffic, the failure is in the TUN layer — device creation, routing or elevation — and the proxy configuration that already works should be left alone. Do not delete a working profile in order to "reset" a TUN problem.

The core starts but pages still fail

A listening port proves the process is alive. It does not prove the server, the rules or DNS are usable. Once the core starts, continue in this order:

  1. Confirm the active server is the one you think it is.
  2. Run a latency test or read the connection log — a result, not a spinner.
  3. Confirm the system proxy is actually applied to the system, not only to the client window.
  4. If an IP address loads and a domain name does not, treat it as a DNS fault.
  5. If only some applications fail, check whether they follow the system proxy at all, and only then evaluate TUN.

Safe recovery and stop conditions

  • Turn TUN off first, then the system proxy, then quit v2rayN normally; confirm a normal HTTPS page loads without any proxy.
  • Keep the last configuration that started successfully as a comparison point.
  • Change one variable per attempt: core version, port, configuration or TUN.
  • Get the client from the official v2rayN releases, and cores only through the client's own update path or the core project's official releases. Do not install a "fixed" core from an unknown mirror, and do not run network reset scripts you cannot read.
  • Stop and hand over if the device is managed, if the process holding the port is unknown to you, or if the same version and error reproduce identically after a clean baseline.

Route by the error string

The dialog is the same for every branch. The table below maps the log line to the page that covers it; each of those pages states its own evidence boundary and marks anything not reproduced.

Log fragmentWhat it points toPage
failed to open geosite.dat / ... > EOFGeo data file missing or truncatedGeo file missing and EOF
ArgumentOutOfRange_CountSubscription content empty, parser ran out of itemsEmpty subscription and argument range
allowInsecure has been removedRetired TLS field inside an imported share linkallowInsecure removed
core IPC pipe did not become ready within 15sService core start timeout, then a retryClash Verge service IPC timeout
nftables operation not supportedTUN came up, a routing rule was refusedsing-box auto_redirect

If your line is not listed, open the error string index and match on the client first, then on the verb.

Frequently asked questions

Does "Failed to run Core" mean my nodes are dead?

No. The core can fail before it contacts any node: a local port still bound by the previous process, a generated field the selected core rejects, missing permissions, or a TUN device that was never created. Read the first core error before you touch the subscription.

Is reinstalling v2rayN the fastest fix?

No. A reinstall removes the log you need and does not resolve a port conflict or a version mismatch. Back up the configuration, keep the log, and change one variable at a time.

I switched servers and it still fails. Is the airport broken?

Not provable from this dialog. If the core never started, every server looks unusable because nothing is listening on the local port. Confirm the core process and the listen port first.

Should I kill the process that holds the port?

Identify it first. Quitting the old client or the old core normally is usually enough. Do not terminate system processes you do not recognise, and hand managed devices to their administrator.

Can I roll back the core version?

Only when a public issue or a release note covers the same version pair, and keep a record of both versions. Fetch builds from official release pages, not from a mirror advertising a "fixed" core.

Sources and verification date

Editorial re-check date: 2026-11-21 (Beijing). This is the date this site checked the public material below; it is not the date each issue was filed, and it does not prove the behaviour still exists on that date.

  • v2rayN public issue reporting failed to listen TCP on 127.0.0.1:10809 with address already in use after TUN was disabled on macOS — one case, port stage.
  • v2rayN public issue reporting a generated DNS configuration rejected by sing-box 1.14.0 under v2rayN 7.24.9, with a maintainer response pointing at a later client release — one version pair, configuration stage.
  • v2rayN public issue reporting a TUN-only start failure on Windows after a sing-box update, with a temporary-rollback suggestion at the time — one case, TUN stage.
  • v2rayN official releases, for the published version list.

This page does not extend a single issue into a rule for all versions, and it does not treat a commenter's manual configuration edit as an official recommendation.

Continue reading

failed to open geosite.dat in v2rayN · ArgumentOutOfRange_Count on subscription update · Full error string index