A dead or flaky network hub almost certainly has Power loss as the reason, faulty cable, faulty port, overheating, or collision overload due to a number of chatterboxy boxes on that port. Hard reset your hub, move to another port to see if that works, check or replace the Uplink, reduce chatter boxes, and upgrade to a switch if necessary.
Symptoms
- All port LEDs dark, or the power LED flickers on and off intermittently
- Every device on the hub loses connectivity at the same moment
- Ping to the gateway shows heavy packet loss, jitter, or duplicate replies
- Throughput collapses under load; small transfers finish, large ones time out
- Hub chassis feels hot to the touch and reboots itself after a few minutes
Common Causes
Power Supply or PSU Brick Failure:
External adapters are more common than the hub. You will get flickering port LEDs, random reboots, and flaky links from a sagging 5 V or 12 V rail, even with the power LED still on.
Damaged cable or worn RJ45 port:
Bent pins, squashed cable jackets and oxidised contacts generate CRC errors and occasional link loss. A single bad port can also generate noise that distorts neighbours.
Collision domain saturation:
Hubs are half duplex and a shared collision domain. Add streaming, backups or VoIP and the collisions increase rapidly, dragging the effective throughput down to well below the rated 10/100 Mbit/sec.
Overheat & ageing capacitor:
a decade old deployed hub, commonly seen bulging electrolytic capacitors, or the vents being dirt and the hub shut down on heat and slow packet loss which is difficult to diagnose.
Broadcast storm or unintentional loop:
Without STP in hubs, if you connect two ports in one switch or connect two switches together with 2 ports from each switch, a loop will be formed and flood the entire network until the network is unusable.
Step-by-Step Fix
- Confirm power delivery and thermal state
Remove the hub's power connector, and then after 30 seconds, plug it back in. Observing the power light, a solid green means OK. If it's flashing and dim, then this indicates a problem with the PSU. Try a different power adapter of the same voltage and polarity. If the hub is very hot, it may be overheating on a carpeted surface or shut in a cabinet – let it cool for a bit, then try again. - Verify cables end-to-end with a known-good patch lead
Check the uplink one first, then check each client one using Cat5e or Cat6 patch leads you know to work, pushing in each RJ45 plug until you hear/feel the retention tab click. Check for any pushed or broken pins inside the RJ45s, and using a cable tester, check each wiremap, the hub will happily continue to operate even if each of your cable runs suffers from a split pair. - Isolate a bad port by moving devices around
Plug a working laptop into port 1, note the LED behaviour, then repeat on every port. A port that never lights, or lights only at 10 Mbit/s when its neighbours negotiate 100, is failing. Mark it with tape and stop using it. If more than two ports are dead, the hub itself is the fault. - Measure the load and count collisions
On a Windows client run netstat -e and watch the Errors and Discards columns over a minute. On Linux use ip -s link show. Rising counters while the hub is idle mean noise or a loop; rising only under load means the collision domain is saturated and you are exceeding what half-duplex can carry. - Break any suspected loop
Trace every cable leaving the hub. Two ports feeding the same upstream switch, or a forgotten patch cable looping back into the hub, will melt the network. Unplug uplinks one at a time; when LEDs stop flashing in unison across all ports, you have found the loop. Never daisy-chain hubs without a plan. - Reduce device count or migrate to a switch
Hubs share bandwidth; switches give each port its own. If you have more than four or five active clients, or anyone doing backups, video, or VoIP, replace the hub with an unmanaged gigabit switch. The hardware cost is minor and the collision problem disappears because switches run full-duplex per port. - Update firmware or retire the unit
A few managed boxes still receive firmware updates; get the exact model number plus the hardware revision number from the base sticker by going to the vendor web page. Generally speaking if the box is about ten years old it is not worthwhile to have it repaired because of bulging capacitors, dried up thermal grease etc. It will be more economical to buy another box and then discard the old shell under WEEE rules.
Quick reference: symptom to likely cause to first action
| Symptom | Likely cause | First action |
|---|---|---|
| Power LED off, no port activity | Failed PSU or dead outlet | Swap adapter, test outlet with another device |
| All ports blinking in perfect sync | Broadcast storm or physical loop | Unplug uplinks one by one to find the loop |
| One port dark, others fine | Burnt port or bad cable | Move client to a spare port, replace cable |
| Links up but heavy packet loss | Collision saturation or CRC errors | Reduce active devices; check netstat -e counters |
| Hub reboots after minutes of use | Overheating or aging capacitors | Improve airflow; plan hardware replacement |
| Slow transfers only during backups | Half-duplex bottleneck | Migrate load to a switched port |
Prevention
- Keep hubs in ventilated space, off carpet, and away from direct sunlight or heat sources
- Label both ends of every patch cable so loops and bad runs are spotted in seconds
- Cap active clients per hub and move high-traffic devices onto a switched port
- Budget a scheduled swap-out for any hub past five years of continuous service
FAQ
How do I differentiate between a hub failure and an ISP outage?
Ping a PC on the same hub (printer or another PC) by local IP. If that fails, something in your LAN or your hub is likely to be the culprit. If local pings work but the internet is dead, the fault is at the router or further upstream at the ISP. This one simple test can save you hours of trial-and-error.
Should I replace my ancient network hub?
Not very often. Most hubs come with inexpensive capacitors and glued-together cases that can be pried apart with extreme difficulty.
Can I simply swap the hub for any switch?
Yes, for the vast majority, especially in smaller deployments. An unmanaged gigabit switch is a convenient, to-the-wall replacement: the same number of RJ45 ports, the same auto-negotiate-quickly-and-flip-on-the-lights feature, and zero configuration. What you get: dedicated bandwidth and full-duplex to each port. If you use port mirroring to connect a packet sniffer, get a managed switch with a SPAN port.