A 3,000-square-meter logistics park in Baiyun District runs 32 Hikvision cameras (DS-2CD3T46 series 4MP bullets on a DS-8632N-K8 NVR). In mid-August the site started losing video: some channels were blank at certain hours, and the screen kept flashing an "insufficient resources" warning. The park manager's first instinct was that the NVR was getting old, and he asked us to quote a replacement 32-channel unit.
After testing on site, our conclusion was the opposite: the NVR was fine. The faults lay in three layers: power, cabling and the hard drive., and a new NVR would not have fixed any of them.
Layer one: PoE overload, and cameras that kept rebooting
The missing channels were concentrated on the west side of the park, nine of them. The NVR's channel log showed those nine cycling online, offline and online again, each outage lasting 30 seconds to two minutes. The cameras were rebooting, not failing.
The power check explained why. The 32 cameras hang off two 24-port PoE switches, each rated at 250W of total PoE output. The west switch carries 18 infrared bullets. With the IR illuminators at full power after dark, each camera draws 8–12W, so demand alone is over 180W. Add two uplinks and one non-compliant PoE splitter, and the real night peak sits just under the 250W ceiling. At that point the switch hits overload protection and drops ports more or less at random. The ports it drops tend to be the longest, heaviest runs, which is why the same channels failed, and only at night..
The fix was an extra 8-port PoE switch (Hikvision DS-3E1310P-SI, RMB 480) to take half the west-side load. With each switch back under 60% utilization, we watched a full night with the IR on and saw no dropouts.
Layer two: oxidized connectors and 30% packet loss
With power fixed, four channels still showed occasional mosaic and short black screens. Pinging those camera IPs returned 12–30% packet loss. All four are outdoor points at the gate and the loading dock, and their RJ45 connectors sat outside the waterproof box.
Opening one was enough: the contacts were black and green with oxidation, and the cable jacket had cracked and let water in. For outdoor surveillance, water and UV are what kill a cable.
We redid all four runs with one-piece waterproof connectors and silicone sealing rings, added corrugated conduit over the jacket, and moved every joint into a waterproof box sealed with glue. Packet loss went to zero.
Layer three: 47 bad sectors and a write queue that could not keep up
Once the picture was stable, the NVR still occasionally reported "insufficient resources, some video lost." That is not a network symptom; it means the NVR cannot write to disk fast enough. The SMART data in the NVR's disk manager told the story. Disk 1, a WD Purple 4TB (WD42PURZ) in service for four years and two months, had 47 reallocated sectors and 11 pending.
A drive with bad sectors makes the head retry again and again. Thirty-two 4MP channels produce a steady write stream of about 45MB/s, and those retries block the write queue, so the NVR has no choice but to discard frames. That is what "insufficient resources" actually means.
We replaced it with two WD Purple 6TB drives (WD63PURZ, surveillance-grade, rated for a 32-channel write load, RMB 1,050 each). The old disk still held 11 days of footage, so we mirrored and exported the critical window first, the week the park needed for a theft dispute, and only then retired it.
What it cost, and who did what
| Item | Detail | Cost |
|---|---|---|
| Power rectification | 8-port PoE switch, RMB 480, plus labor to redistribute the lines | RMB 680 |
| Outdoor cabling | 8 waterproof connectors, 30m of conduit, 4 man-hours | RMB 960 |
| Disk replacement | 2 × WD Purple 6TB (RMB 2,100), footage migration and export, labor | RMB 2,600 |
| Total | — | RMB 4,240 |
Compare the new-NVR route the manager had in mind: RMB 5,800 for the machine plus RMB 1,200 for installation and commissioning, so RMB 7,000, and the power and cabling faults would have stayed exactly where they were. Layered troubleshooting saved the customer close to RMB 3,000 and actually solved the problem.
The order we troubleshoot lost video in, by hit rate
- Start with the NVR event log. Channel up-and-down records point to network or power. "Insufficient resources" points to the disk or the stream configuration. A broken recording schedule points to settings.
- Check disk SMART and recording status. Bad sectors, a full disk without circular overwrite, and an altered recording plan account for about half of all lost video.
- Ping every camera. More than 1% loss means checking the cable and connectors. Intermittent up-and-down means checking PoE.
- Test during the IR hours. PoE overload and aging power supplies both surface once the illuminators switch on.
- Suspect the NVR itself last. A genuinely failing NVR takes every channel down at once or refuses to boot. Single-channel or time-window losses almost never trace back to it.
ACCPC maintains CCTV systems across Guangzhou. An annual contract for a park or shop with up to 32 channels runs RMB 3,600–6,000, covering quarterly inspections, connector service, disk health monitoring, unlimited remote support and on-site labor at no extra charge. A single on-site inspection starts at RMB 300. We also wholesale Hikvision, Dahua and TP-LINK equipment in Guangzhou at channel prices, and integrators are welcome to ask. Call 020-39029800.