Detecting and Resolving IP Address Conflicts
Two devices, one address: the network's version of identity theft. An IP address is supposed to mean exactly one machine. Nearly everything a network does is built on that assumption. So when two devices both claim 10.1.20.31, the network does not crash honestly. It does something far worse: it half-works. Connections succeed, then die mid-session. A server is reachable from one desk and invisible from the next. A ping comes back cheerfully — from the wrong machine. Address conflicts produce the weirdest intermittent symptoms in networking. That is why they routinely burn hours of troubleshooting aimed at everything except the actual cause.
There is good news. Once you understand the one mechanism underneath the weirdness, conflicts stop being spooky. They start being a tidy, almost enjoyable hunt with a clear ending. This guide covers how conflicts happen in the first place (the origin stories repeat — you will recognize yours). It covers why the symptoms flicker the way they do and how to detect a conflict deliberately instead of by rumor. And it gives a step-by-step drill for finding both claimants and deciding who keeps the address. The drill also makes sure this particular fight never happens again.
In this guide
- How two devices end up with one address
- Why the symptoms are so weird
- Detection: the tells and the tools
- The manual confirm, step by step
- The resolution drill
- Prevention wiring
- The conflict response runbook
How two devices end up with one address
Conflicts are never spontaneous — somewhere, a process failed. The origin catalog is short and depressingly familiar:
- The unrecorded static inside the DHCP pool. The classic. Someone hand-types a static address that happens to sit inside the range the DHCP server is leasing from, and records it nowhere. The server, entirely innocent, eventually leases that address to a laptop. This is not bad luck; it is a design sin. Statics must live outside pool ranges, in a band reserved for them. That is exactly the reservation policy laid out in IPAM basics.
- The returning device. A laptop closes its lid and spends three weeks in a drawer. Its DHCP lease expires; the address is leased to someone else. The laptop wakes up confident the address is still its own and starts using it. It is a guest returning at midnight to a room the hotel re-let, still holding the old key. Modern systems usually re-check on waking, but "usually" is carrying weight in that sentence.
- The cloned VM. Copy a virtual machine and you copy its configuration — including a baked-in static address. Boot the clone on the same network as the original, and the twins fight over one identity.
- The printer-installer's favorite trick. A vendor tech arrives, assigns the new printer or door controller a static address chosen by vibes, tells no one, and departs. Whether it collides today or sits as a landmine for a year is pure chance. Vendor techs are lovely people who should never be left alone with an address space.
- The rogue DHCP server. Someone plugs in a home router the wrong way round, or enables DHCP on a lab box. Now a second server is handing out addresses from an overlapping range. It is manufacturing conflicts wholesale rather than retail, with a side order of devices receiving the wrong gateway entirely.
- The multihomed mistake. A second network card configured with the first one's address, a typo transposing .12 into .21, a restore of an old config onto new hardware. Fat fingers remain undefeated.
Why the symptoms are so weird
To see why conflicts flicker instead of failing cleanly, you need one piece of machinery: ARP, the Address Resolution Protocol. On a local network, devices do not actually deliver traffic to IP addresses. They deliver it to hardware MAC addresses, the factory-burned identifier of each network card. ARP is the lookup between the two. A device shouts to the whole segment, "who has 10.1.20.31?" The owner replies "me — here is my MAC." The asker files that answer in a short-lived cache. It sends everything for 10.1.20.31 to that MAC until the cache entry ages out, typically after a few minutes.
Now put two claimants on the segment. Both hear the question; both answer truthfully by their own lights; the asker keeps whichever answer lands last. From that moment, this particular asker sends all traffic for the address to that particular claimant. That lasts until the asker's cache expires. The asker asks again, and possibly gets the other answer. Multiply by every device on the subnet, each holding its own cache, each refreshing on its own schedule. You get the signature symptom pattern: works-then-doesn't, one victim at a time. Your desktop reaches the real server (its cache points at the right MAC) while your neighbor's session dies mid-transfer (theirs just flipped to the impostor). Five minutes later the roles swap. Nothing is down; everything is unreliable; the ticket queue fills with contradictions, and everyone's screenshots are simultaneously accurate.
Watch one conflict unfold on a clock and the whole pattern snaps into focus. 9:02 — a returning laptop wakes and announces its stale claim on the file server's address. Every cache on the segment updates obediently. 9:03 — the accounting team's mapped drives freeze mid-save. The server's own logs show nothing, because nothing reaches it to be logged. 9:07 — the laptop's owner, finding "the Wi-Fi slow," wanders to a meeting room. The laptop roams. The server's next broadcast reclaims some caches, and half the victims recover spontaneously. 9:20 — the laptop reconnects and re-announces; the outage returns in a different corner of the floor. By 9:40 there are nine tickets describing four contradictory problems, and whichever machine the helpdesk tests from happens to work. Nothing in that half hour was random. It was ARP caches flipping on their individual timers. But without the mental model, it is indistinguishable from haunting.
It also explains the eeriest tell: pings answered by the wrong machine. The address responds beautifully — because the printer now wearing the server's address is in perfect health. And there is a cruel asymmetry in who suffers. Devices announce their claim to an address when they bring it up, using a broadcast called a gratuitous ARP. It says "attention everyone: 10.1.20.31 is at this MAC now" and rewrites listeners' caches. So the second device to speak often wins visibility. Its announcement stamps its MAC over every cache on the segment. Traffic flows to the newcomer. The original owner — the server that was there first, doing nothing wrong — goes quietly deaf. Its incoming connections evaporate without a single error in its own logs. The victim is silent and the culprit looks fine. Networking has a dark sense of humor.
One more variable sets the blast radius: which address is being fought over. Two laptops sharing an address inconvenience two people. A conflict on a server address takes down a service for everyone, in the flickering pattern above. And a conflict on the default gateway's address — the router everyone's traffic leaves through — is the five-alarm variant. Every device whose cache flips to the impostor loses the outside world entirely. So the whole subnet strobes on and off in patches. If a conflict report involves the gateway, skip the queue and treat it as a full outage in progress.
Detection: the tells and the tools
You can meet conflicts two ways: by rumor ("the ERP is being weird again") or by instrumentation. The instrumented ways, from cheapest to fanciest. First, the operating systems already know. Before using an address, a well-behaved device probes for it with ARP. If someone answers, it complains. Windows pops its duplicate-address warning and writes an event to the system log. Other operating systems log equivalent gripes. Those log entries are gold lying on the ground. If your log collection or Windows-focused monitoring gathers system events centrally, a search for address-conflict entries turns anecdotes into timestamps and MAC addresses. The Windows Network Monitoring guides cover getting at Windows' view of the network generally.
Second, listen for competing claims. Since every device announces address claims via gratuitous ARP, anything positioned to listen can notice two different MACs claiming one address. Some switches can flag or block suspicious claims. Monitoring or security tools with an ARP-watching feature raise exactly this alert. If your gear offers duplicate-address detection, turn it on; it is the difference between the network telling you and users telling you.
Third — available on any network, with no features required — the ARP-table sweep. Your routers and switches maintain ARP tables: their current beliefs about which IP maps to which MAC. Harvest those tables on a schedule, keep the history, and diff. The same IP appearing with different MACs across snapshots, or in two devices' tables at once, is a conflict announcing itself in writing. As a sketch:
# Scheduled ARP harvest, the poor-admin's conflict detector: # 1. every 15 min, pull ARP tables from the core switch/router # (SNMP, or a scripted login and "show arp" / "show ip arp") # 2. store as date-stamped snapshots: IP, MAC, source device # 3. diff against history: # same IP + new MAC = lease churn (normal) OR a conflict # same IP + two MACs in the same window = conflict, alert # 4. keep 90 days — the history also answers "when did it start?"
The DHCP angle helps too. A lease table entry and a sighted static with the same address but different MACs is the pool-collision conflict caught red-handed. That is one more payoff of the reconciliation habit from IPAM basics.
The manual confirm, step by step
However the suspicion arrives, confirm it the same humble way — from any machine on the affected subnet, with two built-in commands. The trick is to catch the address wearing two faces:
ping 10.1.20.31
arp -a 10.1.20.31 :: note the MAC that answered, e.g. 00-1b-a9-...
:: now power off or unplug the DEVICE YOU BELIEVE owns the address,
:: then clear the cached entry (admin prompt) and ask again:
arp -d 10.1.20.31
ping 10.1.20.31
arp -a 10.1.20.31 :: still answering, with a DIFFERENT MAC?
:: two claimants confirmed.
Read it like a detective: if the rightful owner is off and the address still answers, someone else is home. In that case, the second MAC is their fingerprint. Run the same check from two different machines simultaneously and you will sometimes catch the flip-flop live. You see two askers, the same IP, and two different MACs cached at the same moment. Save both MACs; the whole resolution phase runs on them. If shutting down the presumed owner is not an option (it is the production server, after all), work the other direction. In that case, compare the answering MAC against the MAC recorded for the server in your register or on the machine itself. A mismatch is the same confession.
Tip: a MAC address tells on its maker. The first half of a MAC is a manufacturer prefix (the OUI), and free lookup tables translate it. So before any physical hunting, thirty seconds of lookup often converts "unknown second claimant" into "some brand of printer." That shortens the suspect list dramatically. Half the time the interloper turns out to be a printer. It is always a little satisfying when it is.
The resolution drill
Confirmed conflict in hand, run the drill. If the victim is something important, buy calm first with a containment move. Once you know the interloper's MAC and its switch port (next paragraph), administratively shutting that port ends the flip-flopping immediately. At that point, simply unplugging the device also works, if it is the sort of interloper you can walk to. The rightful owner re-announces, caches heal within minutes, and you can finish the investigation without a live audience. Containment is not resolution. The interloper still believes it owns the address and will resume the fight the moment it reconnects. But containment converts an active outage into a background task, which is a fine trade.
Find both claimants physically. A MAC becomes a location through your switches. Every managed switch keeps a MAC address table listing which MAC was last seen on which port. The table is typically shown with a command in the family of show mac address-table. Look up each claimant's MAC on the core switch. If the port points at another switch, follow the trail downstream and repeat. When you reach a port with a single MAC on it, that cable ends at your device. Wall jack, patch panel, done — this port-tracing walk is the standard "MAC to human" technique and works for any mystery device, conflicted or not.
Decide the rightful owner. This is a records question, not a shouting match. The assignment register from IPAM basics says who the address belongs to, and the register wins. The server documented at .31 keeps .31. The visiting laptop, the cloned VM, or the freelancing printer is the interloper. That is true even if the interloper spoke second and currently "owns" every ARP cache on the floor. No register entry on either side? Then you have found two process failures for the price of one; pick the owner deliberately and write both down.
Fix the interloper, not just the symptom. If it should have been a DHCP client all along, return it to DHCP (on Windows, ipconfig /release then /renew). Then it will receive a legitimate lease and peace resumes. If it genuinely needs a fixed address, assign it a proper one from the correct reserved band, recorded in the register at assignment time. Verify the peace: after the fix, caches settle as they expire or hear the rightful owner's next announcement. Give it minutes. Clear ARP entries on the loudest victims if you are impatient. Then confirm the address answers from the right MAC everywhere and the silent victim's services are reachable again. And record the incident: one line in a conflict log — date, address, both MACs, cause, fix. That log is more than paperwork: repeat offenders reveal process gaps. Three conflicts traced to vendor installs is not three accidents; it is a missing rule about vendor visits.
Prevention wiring
Every origin story earlier has a structural counter, and they are all cheap. Statics live outside pool ranges, full stop — the single rule that retires the most common conflict outright. Prefer reservations over hand-typed statics where your DHCP server supports them. A reservation tells the server "this MAC always gets this address." That gives the device a stable identity while keeping the assignment centrally recorded, centrally visible, and impossible to fat-finger on the device. The register and the enforcement mechanism become the same thing. No static without a register entry, ever — the intake rule, sixty seconds at assignment time, most of this article avoided. And treat the known conflict-prone moments with ritual awareness. Cloning a VM gets a checklist line ("network config regenerated before first boot on a shared segment"). Vendor visits get an escort policy. The tech is welcome, and the address they need is issued from your register, not invented from theirs.
The conflict response runbook
The whole guide as one table — print it, or at least know where it lives, because conflict days are flustered days:
| Stage | What you do | Evidence in hand |
|---|---|---|
| Symptom | Recognize the pattern: intermittent, works-for-me-not-for-you, one subnet, pings answered oddly, duplicate-address warnings in logs | Ticket cluster; log events |
| Confirm | Ping + arp -a; owner off, address still answers; catch the second MAC | Two MACs, one address |
| Trace | MAC address tables, core outward, to a final port per claimant; OUI lookup for make | Two physical devices |
| Resolve | Register decides the owner; interloper re-DHCPed or given a proper reserved static; verify caches settle and victim recovers | Clean ARP everywhere |
| Prevent | Log the incident; fix the process that allowed it — pool borders, intake rule, cloning checklist, vendor escort | One line in the conflict log |
Where to go next
Conflicts stop being gremlins the moment you can see ARP: two claimants, flip-flopping caches, a silent victim — then a confirm, a trace, a register ruling, and a one-line log entry. The deeper cure is everything around this guide. The registers and reservation policy in IPAM basics make conflicts structurally rare. Healthy pool borders in DHCP scope monitoring keep the server from ever leasing a taken address. The rest of the IP Address Management & Subnet Monitoring guides build the address space where identity theft has nowhere to start.
Put this into practice with SysPeep Monitor
A conflict hunt is mostly a hunt for what is actually on the subnet. That part can be standing ready before anything breaks. SysPeep Monitor auto-discovers the devices on your network and keeps a live picture of them in one web console. So when an address starts acting strange you begin with facts instead of folklore. The free trial is a quick install on any Windows machine.
Download Free TrialLearn About SysPeep