Guides, Nov 17, 2026

# How to check if your VPN server IP is blocked

To check if a VPN server's IP is blocked, test it from inside the country or network where users fail, and compare that with a test from outside at the same time. If the server works from abroad but connections fail from inside, it is a block, not an outage. Then test DNS, the IP, the TLS name and the protocol one at a time to see which of them is blocked. OONI Probe, RIPE Atlas and users who agree to run a short test are the practical ways to get that inside view.

Tahir Mehmood, 10 min read

This post is for people who operate VPN and proxy servers and get the message "it stopped working for me" from users in one country while every dashboard is green. It covers how to find out what is happening on the users' side of the network, with tools that publish their methods. One warning first: running network tests can carry legal risk in some countries, for you and for anyone who runs a test for you. I come back to that below.

## How does a blocked VPN server look from the client?

A block looks like a timeout, a reset or a slow connection, and the exact symptom tells you what is being blocked. The same server can look perfectly healthy from outside the country, because the block happens on the users' network, not on your server.

| What is blocked | What the client sees | Quick test from inside |
| --- | --- | --- |
| DNS name | The hostname resolves to a wrong address, or not at all | `dig +short vpn.yourdomain.com` |
| IP address | TCP connections time out on every port, ping may fail too | `nc -vz -w 5 176.61.144.20 443` |
| TLS name (SNI) | TCP connects, then the connection resets or stalls right after the TLS ClientHello | TLS to the same IP with your name and with another name |
| Protocol | TCP and TLS work, but the VPN handshake fails, or the tunnel works for a short time and then stops | Run the real client and read its log |
| UDP or QUIC | WireGuard, Hysteria2 and TUIC fail, TCP protocols on the same server work | Compare a UDP protocol with a TCP one |
| Speed | Everything connects, but transfers are very slow | A timed download through the tunnel and without it |

Real blocks often combine these. A network may block an IP for one protocol and slow down everything else. That is why I test one layer at a time instead of only asking "does it work?".

## How do I tell a block from an outage?

Compare a test from inside the affected network with a test from outside, run in the same few minutes. If outside fails too, it is your server or your provider, and you debug it as an outage. If outside works and inside fails, something on the path to those users is in the way.

The pattern across users narrows it further:

-   **One user fails, others on the same ISP work:** their own network, device or client config.
-   **Every user on one ISP fails, other ISPs in the country work:** a block or a routing problem at that ISP.
-   **Every ISP in the country fails, users abroad work:** a national-level block, or a problem between that country and your hosting provider.
-   **The server worked for days and then failed for one country only:** a new block is the likely cause. A server outage doesn't pick one country.

Before you call it a block, rule out the boring causes that also hit one network only: MTU problems on mobile networks and PPPoE lines, which I covered in [WireGuard MTU problems](https://tunnelhq.com/blog/wireguard-mtu-problems/), and an expired or wrong certificate that one client app validates more strictly than others.

## How do I test the IP, the domain and the protocol separately?

Give someone inside the network a short list of commands that test one layer each, in order, and ask for the full output. Each step only means something if the step before it passed. On Linux and macOS:

```
# 1. DNS: does the name resolve to your server's address?
dig +short vpn.yourdomain.com

# 2. IP: can it open a TCP connection to the address at all?
nc -vz -w 5 176.61.144.20 443
nc -vz -w 5 176.61.144.20 22    # or any other port you know is open

# 3. TLS name: the same IP, first with your name, then with another name
curl -v -k --connect-timeout 5 --resolve vpn.yourdomain.com:443:176.61.144.20 https://vpn.yourdomain.com/
curl -v -k --connect-timeout 5 --resolve other.yourdomain.com:443:176.61.144.20 https://other.yourdomain.com/

# 4. Path: where do TCP packets to port 443 stop?
mtr -r -T -P 443 -c 20 176.61.144.20
```

`--resolve` makes curl connect to the IP you give it while it sends the name you give it, so step 3 changes only the TLS name and keeps everything else the same. `-k` skips certificate checks, because here you only care whether the network lets the handshake through. Step 3 only applies to a server that speaks TLS on that port.

How I read the results:

| Result | Likely meaning |
| --- | --- |
| `dig` returns another address, or nothing | DNS tampering on the users' resolver. The IP itself may still be reachable |
| `nc` times out on every port, including 22 | The IP is blocked |
| `nc` works, curl with your name is reset, curl with the other name gets a TLS answer | The filter reads the SNI, and your name is on its list |
| Both curl calls get a TLS answer, but the VPN client fails | The protocol is being detected, or the server rejects the client. Compare with a client outside |
| Everything connects, but slowly | Throttling. Measure it before you change anything |

For UDP protocols, `nc` tells you nothing, for the same reason a UDP port check is useless from outside: a healthy server sends nothing back to a packet it can't authenticate. The only UDP test is the real client. If WireGuard or Hysteria2 fails while a TCP protocol on the same server works from the same phone, UDP or QUIC is being filtered on that network.

For the protocol layer, ask for the client log, not a screenshot of the error. Clients such as v2rayN, v2rayNG and NekoBox show a log. A tunnel that comes up and dies after a few seconds or a few megabytes is a different problem from a handshake that never completes. If you run REALITY, [Xray REALITY not working](https://tunnelhq.com/blog/xray-reality-not-working/) covers the server-side causes worth ruling out before you blame the network.

## How do I use OONI Probe and OONI Explorer to check a block?

OONI Probe tests whether a URL is reachable from the network it runs on, and OONI publishes every measurement as open data, so you can search OONI Explorer or the OONI API for tests of your domain by country. The Web Connectivity test performs a DNS lookup, a TCP connect and an HTTP or HTTPS request over the user's network, does the same from a control server, and flags a difference as an anomaly. It classifies the difference as DNS, TCP/IP or HTTP blocking.

Someone inside the country can test your server's hostname with the [OONI Probe CLI](https://ooni.org/support/ooni-probe-cli/):

```
ooniprobe run websites --input https://vpn.yourdomain.com/
ooniprobe list
```

OONI Run links do the same from the mobile app, for people who won't use a terminal. You can then read the results from anywhere through the OONI API:

```
curl 'https://api.ooni.io/api/v1/measurements?probe_cc=IR&domain=vpn.yourdomain.com&test_name=web_connectivity&since=2026-11-01&limit=50'
```

Each result has `anomaly`, `confirmed` and `failure` fields and a link to the raw measurement. An anomaly is a reason to look closer, not proof: read the raw measurement to see which step failed.

Know the limits before you use it:

-   **It tests a website, not your VPN protocol.** It is useful when your server answers HTTPS on 443 (a Trojan or VLESS server with a web fallback, for example). For a WireGuard-only server it tells you little.
-   **Results are public.** Your hostname, the tester's country and network (ASN) and the time end up in open data. Don't test a hostname you want to keep private.
-   **Running it has risks.** OONI's own [risks page](https://ooni.org/about/risks/) says anyone monitoring the network can see that OONI Probe is running, and that in some countries using it may lead to prosecution. Never ask a user to run it without telling them that.

OONI Explorer is also worth searching without running anything. If other measurements show a wave of TLS failures for many domains on one ISP in the same week, your server is probably not the only target.

## How do I use RIPE Atlas to test from inside a country?

RIPE Atlas lets you run ping, traceroute (including TCP traceroute to a port), DNS and TLS measurements from volunteer-hosted probes, and you choose probes by country or network. HTTP measurements are restricted, so you can't fetch a page through it, and it can't run a VPN client. You pay for measurements with credits, which you earn by hosting a probe (RIPE NCC members can also run some for free). With [ripe-atlas-tools](https://github.com/RIPE-NCC/ripe-atlas-tools):

```
pip install ripe.atlas.tools
ripe-atlas configure --set authorisation.create=<your-api-key>

# Is the IP reachable at all?
ripe-atlas measure ping --target 176.61.144.20 --from-country ir --probes 10

# Do TCP packets to 443 get through, and where do they stop?
ripe-atlas measure traceroute --target 176.61.144.20 --protocol TCP --port 443 --from-country ir --probes 10

# What do local resolvers return for your name? No --target means each probe uses its own resolver.
ripe-atlas measure dns --query-argument vpn.yourdomain.com --from-country ir --probes 10

# Does a TLS handshake complete, and which certificate comes back?
ripe-atlas measure sslcert --target vpn.yourdomain.com --port 443 --from-country ir --probes 10
```

Replace `ir` with the two-letter code of the country you care about. Per the [RIPE Atlas documentation](https://atlas.ripe.net/docs/getting-started/user-defined-measurements.html), one-off measurements return results within seconds, which makes them good for "is it blocked right now?".

The limits:

-   **Coverage varies a lot by country.** Look at the probe map for your target country before you plan around it. Most probes sit on fixed home, office and data center connections, so mobile networks are rarely covered.
-   **No UDP handshake test.** Ping and TCP traceroute tell you about the IP and the path, not whether WireGuard or QUIC gets through.
-   **Results are public by default.** The API's `is_public` field defaults to `true`.

## Can I ask users to run the tests for me?

Yes, and for protocol-level blocks it is often the only option, because no public measurement network runs your VPN client. Keep it small and transparent: a short script with the commands above, output they can read before they send it, and nothing that collects more than you need (no device identifiers, no browsing data). Ask them which ISP and connection type they use (mobile or fixed), because the answer is often different between the two.

Some operators keep a small machine on a residential or mobile connection inside the country to test from. Whether that is legal and safe depends entirely on the country, so treat it as a legal question first and a technical one second.

## What should I do once I know what is blocked?

Match the fix to the layer that failed, and change one thing at a time so you know what worked. If only DNS is tampered with, clients that connect by IP or use encrypted DNS will work again. If the IP is blocked, a new IP helps, but an IP from the same provider range may be blocked again soon. If the TLS name is blocked, change the name. If the protocol is detected, a different protocol or transport is the real fix, and changing the IP won't help.

Whatever you change, test it from inside again before you send it to every user. A new config that works from your office proves nothing about the network that blocked the old one.

## What can a monitoring service see, and what can't it?

A monitoring service tests from its own checkers, which sit in data centers, so it can tell you whether a server works from each location it tests from. It can't see a block that only exists inside a country it doesn't test from. That split is still useful: if the service says the server is healthy everywhere it looks while users in one country fail, you have ruled out the server and can focus on the path.

Full disclosure: TunnelHQ is one of these services. It completes the real handshake from its checkers, and on higher plans you choose the [test countries](https://tunnelhq.com/pricing/) for each server. Each location gets its own result, and a server that is up in one location and down in another shows as Partial instead of Down. TunnelHQ does not test from inside Iran, China or Russia, so for blocks there, the in-country methods above are the ones that work.

Source: https://tunnelhq.com/blog/check-vpn-server-ip-blocked/
