The port forward worked for months. The service behind it was reachable from outside, you stopped thinking about it, and then one day it was just dead, or stranger still, the same rule started answering with a completely different device. You open the router, and the port-forward rule is exactly as you left it: same external port, same internal address, same everything. Nothing changed. That’s the problem. The rule points at an address, and the address moved.
The reason this is so disorienting is that the one thing you’d check, the rule, is innocent. It says 203.0.113.10:8080 -> 192.168.1.50:8080 today exactly as it did the day you created it. Every byte of the configuration you’d think to inspect is correct. The part that broke is invisible from that screen: which device is actually sitting at 192.168.1.50 right now.
Table of Contents
A port forward points at an address, not a device
When you set up port forwarding, you tell the router: traffic arriving on this external port should go to this internal address and port. You typed an IP, 192.168.1.50, because that’s what the device had when you set it up. From then on, the rule is a pointer to that number. It has no idea which machine is there; it knows the address and nothing else.
That’s fine as long as the device keeps the address. The trouble is that on a default home network, it was never promised the address in the first place. It was lent one.
Why the address moves: the DHCP lease
Your router runs a DHCP server, and by default it hands out addresses from a pool, say 192.168.1.100 through 192.168.1.200, to any device that asks. Each assignment is a lease: the address is yours for a set time (often a day or a week), after which the device must renew it. The DHCP protocol (RFC 2131) is explicit that a lease is temporary and that the server may hand back a different address on renewal; a device is only ever borrowing.
Most of the time a device renews and keeps the same address, which is exactly what lulls you. But any of a dozen ordinary events can break the streak: the device reboots, the router reboots, the device sleeps past its lease and drops off, the pool fills up, or the network simply reshuffles. When the device comes back and asks for an address, the router gives it an address from the pool, not necessarily the old one. Now your server is at 192.168.1.73, and your port-forward rule is still faithfully pointing at 192.168.1.50.
The port-forward rule didn’t change. The lease under it did, and the lease is the part you can’t see from the forwarding screen.
The rule sends external traffic on port 8080 to 192.168.1.50 – the address the device had that day.
The lease lapses or the device reconnects. DHCP hands it a new address, 192.168.1.73, from the pool.
Traffic arrives at the old address and finds nobody – or finds whatever device now holds 192.168.1.50.
The rule never changed, the lease under it did. A DHCP reservation pins the device to one address, so the forward's target stops moving.
The tell: the rule is unchanged, the device’s IP isn’t
The diagnosis is a single comparison: what address does the rule target, and what address does the device actually have right now? Read them side by side and the gap jumps out.
# On the server itself - what address does it ACTUALLY have right now?
ip addr show # Linux / macOS: look for the inet 192.168.x line
ipconfig # Windows: the IPv4 Address line
# 192.168.1.73 <- the device's real address today
Code language: Bash (bash)
Then open the router’s port-forward page and read the internal address the rule targets:
203.0.113.10 : 8080 -> 192.168.1.50 : 8080 <- the rule still says .50Code language: CSS (css)
.73 on the device, .50 in the rule. That mismatch is the bug, in full view, the moment you put the two screens next to each other. It’s the same shape as any defect where the thing works on your machine but fails where it actually runs: the configuration you’re staring at is correct for a world that has quietly moved on.
Why it broke on a day you changed nothing
The maddening part is the timing: you didn’t touch the router, didn’t touch the server, and it broke anyway. That’s because the trigger isn’t a change you made; it’s the lease clock. A power blink that reboots the house, a router firmware update that restarts DHCP, the server sleeping through a weekend, a new phone or guest laptop joining and claiming a freed address, any of these can hand your server a new lease. The port-forward rule has been a time bomb since the day you created it; the lease simply hadn’t rolled yet.
And there’s a nastier version. If the device that used to be at 192.168.1.50 lost its lease and some other device later grabbed .50 from the pool, your port forward now points at the wrong machine. External traffic meant for your service is being delivered to a random laptop or a smart plug that happens to hold the address today. It won’t answer the way your service does, so the symptom looks like a broken or misbehaving service, when really the packets are arriving at a stranger. A reassuring “the rule is still there” is exactly the kind of signal that reports on the thing it checked and nothing more: the rule’s existence says nothing about who’s behind the address.
Fix: pin the address with a DHCP reservation
The cure is to make the target address stop moving, so a rule that points at it stays correct forever. Don’t edit the port forward; anchor the device.
The right fix for almost everyone: a DHCP reservation. In the router’s DHCP settings, bind the device’s MAC address (its hardware ID, which never changes) to a specific IP. From then on, every time that device asks for an address, the DHCP server hands it the same one, lease after lease, reboot after reboot. Your port-forward rule now points at an address the device is guaranteed to keep.
# In the router's DHCP / LAN settings, add a reservation:
# Device MAC aa:bb:cc:dd:ee:ff -> reserve 192.168.1.50
# Then reboot the device (or renew its lease) so it picks up the reserved address.Code language: PHP (php)
Two practical notes that save a second round of confusion:
- Reserve an address that matches your rule, or update the rule to match the reservation. If you reserve
.50but the device is currently on.73, renew its lease (reconnect it, or reboot it) so it actually moves to.50before you trust the forward again. Reserve first, verify the device landed there, then rely on the rule. - A static IP set on the device also works, but it must sit outside the DHCP pool. If you hard-code
192.168.1.50on the device while the router’s pool still includes.50, the router can lease.50to something else and you’ve created an address conflict, two devices, one address, intermittent failures for both. A DHCP reservation avoids this because the router is the one authority handing out addresses; it won’t lease a reserved one to anyone else. Prefer the reservation.
After you pin it, re-check from outside that the service answers, and you’re done, permanently this time, because the one thing that was drifting is now nailed down.
The short version
A port forward points at a LAN IP address, not at a device. On a default home network that address came from DHCP as a temporary lease, so the router is free to hand the device a different address when the lease lapses or the device reconnects. When that happens your rule still points at the old address, so the service goes dark, or, if another device has since grabbed the old address, your forward delivers traffic to the wrong machine. The rule itself never changed; the address under it did. Fix it by pinning the device with a DHCP reservation (bind its MAC to a fixed IP), so the target the rule points at can never move again.
Frequently Asked Questions
Why did my port forward stop working when I didn’t change anything?
Because the thing that changed wasn’t the rule, it was the device’s IP address. Your port forward points at a specific LAN address, and that address was almost certainly assigned by DHCP as a temporary lease. When the lease lapsed or the device reconnected (after a reboot, a power cut, a router update, or the device sleeping too long), the router handed it a different address, and your rule kept pointing at the old one. Nothing you did broke it; the lease clock did. Pin the device with a DHCP reservation and the address stops moving.
What’s the difference between a DHCP reservation and a static IP?
Both give a device a fixed address, but they set it in different places. A DHCP reservation is configured on the router: you bind the device’s MAC address to a chosen IP, and the DHCP server always hands that device the same address. A static IP is configured on the device itself, which stops asking DHCP at all. The reservation is safer for most people because the router remains the single authority on who gets which address, so it won’t accidentally lease your “static” address to another device. If you do set a static IP on the device, make sure it’s outside the router’s DHCP pool to avoid exactly that conflict.
Could my port forward be sending traffic to the wrong device?
Yes, and it’s the sneakier failure. If the device that held the forwarded address lost its lease, and another device later picked up that same address from the DHCP pool, your port-forward rule now delivers external traffic to whatever machine currently holds the address, a laptop, a phone, a smart-home gadget. The service looks broken because the packets are arriving at a device that doesn’t speak its protocol. A DHCP reservation prevents this by guaranteeing the intended device always holds the address the rule targets.
How do I confirm this is the problem?
Compare two numbers. On the server, read its current address (ip addr show on Linux or macOS, ipconfig on Windows) and note the 192.168.x value. Then open the router’s port-forward page and read the internal address the rule targets. If they differ, that’s the whole bug, visible in one glance. The device has drifted to a new address and the rule is still pointing at the old one.