Rescued a water-damaged Cisco CBS350-24T-4G with a jury-rigged ATX PSU, then spent a full day chasing a phantom packet-loss bug that turned out to be the switch itself
A while back I got my hands on a Cisco Business CBS350-24T-4G that had taken a direct hit of water. The internal PSU was completely fried, burnt board, dead on arrival. Opened it up, confirmed the mainboard itself was untouched, so I rigged an external ATX PSU with the classic paperclip PS_ON trick to bring it back to life.
First boot: the System LED just blinked forever, boot loop. Turned out the config was corrupted from the power event. After a few power cycles it silently fell back to a factory/golden image and came up clean. Recovered full web management access, renamed all the ports, set static IPs, tightened up VLAN/STP settings, the works. The switch became the core of my home network: router uplink, home server, gaming PC, TV, all fanned out from it.
Weeks later, random and intermittent packet loss started showing up during online gaming (Valorant), sometimes hitting 70 to 80% spikes for a couple seconds, then clean for a while, completely unpredictable.
Spent a full day tracking it down methodically:
- Ruled out the ISP/ONT (checked optical RX power, hardware self-test, firmware, all clean)
- Ruled out the PC's NIC power-saving settings (fixed a real issue there, loss dropped from around 6 to 7% down to around 2%, but didn't fully go away)
- Ruled out Spanning Tree topology changes (log showed nothing recent)
- Ruled out CRC/physical errors (RMON stats on both relevant ports: 0 errors across 500M+ packets each)
- Ruled out CPU load (switch sitting under 5%)
- Ruled out EEE/Green Ethernet interop weirdness (disabled it, no change)
- Even measured the jury-rigged PSU's 12V rail live under load with a multimeter, rock solid at 12.2V, barely any ripple
Final test: bypassed the switch completely and plugged the PC straight into the router with the exact same cable. Ran it for 8+ minutes. Zero packet loss.
Put the switch back in the path, same cable, same everything: packet loss came right back, consistently around 2 to 4%.
So it really is the switch itself, most likely some lingering hardware quirk from the whole water/firmware-corruption saga, invisible to every counter and log the switch exposes, but very real and reproducible for actual forwarded traffic.
Ended up buying a cheap $30 unmanaged 8-port TP-Link switch to replace it for the gaming PC's path. Zero loss since. Attached a ping graph as proof of the bug (8 min continuous ping through the Cisco, around 4% loss with visible timeout spikes), good closure for what became a pretty deep rabbit hole for a switch I got for free.
Lesson learned: sometimes a "fully recovered" piece of enterprise gear can look 100% healthy on paper (clean RMON, clean CPU, clean STP, stable PSU) and still have something wrong deep enough that only a full bypass test reveals it.
