The camera you need is usually the one that stopped recording in March

How a camera stops recording without telling you

The camera itself is rarely the problem. What actually happens:

• The power over Ethernet port on the switch fails, or the switch is over its power budget after somebody added two cameras, and the last device to boot never comes back.
• A drive in the recorder fails. On a single-drive recorder that is everything. On a multi-drive recorder it may be one camera’s worth of footage, quietly.
• The camera reboots in a loop from a bad power supply or water in a connector, so it appears online for 40 seconds at a time.
• Recording is set to motion only and the motion zone was never adjusted, so the camera records the tree and not the doorway.
• A firmware update or a password change breaks the recorder’s connection to that channel. Live view through the camera’s own app still works, which is why nobody notices.
• The disk is full of one camera’s footage. High resolution and constant motion on one channel can eat the shared pool and shorten retention on every other camera.

None of these announce themselves. Most recorders can email or notify on video loss and disk error, and on most sites that feature has never been switched on.

Retention drift, and why the setting is not the number

The recorder is set to 30 days. The question is what is actually on the disk.

Retention gets eaten by anything that increases how much data is written: cameras added after the original design, resolution or frame rate turned up during a callout, a new light that makes a night scene busy enough to trigger constant motion recording. Retention is also cut short by a failed drive in a pool, because the pool is smaller now.

The check takes two minutes and no tools. Open playback, go to each camera, and drag the timeline back to the oldest date that returns video. Do it per camera, not once for the system, because they frequently differ. Write the dates down. That list, against the number you thought you had, is the only retention report that means anything.

Focus drift, spiderwebs and the rest of the physical world

Cameras live outside and the world works on them.

Varifocal lenses drift over years of heating and cooling, so a scene set up sharp at install goes soft. Dome bubbles haze and yellow under UV, which drops contrast and scatters infrared at night. Condensation gets inside housings once a seal ages, and the first sign is a soft picture in the morning that clears by noon.

Spiderwebs are the most common and most annoying one. A web across the lens catches infrared at night, glows white, and either blinds the camera or triggers motion recording all night, which is what quietly destroys your retention.

Wasp nests inside housings, cobwebs behind the bubble, tree growth over a summer, a delivery driver knocking a camera off aim with a pallet, and the seasonal change in sun angle that puts glare into a lens for three weeks in autumn. All of these are found by looking, and none are found by checking the recorder.

Firmware, passwords and the network side

Cameras are computers on your network, and neglected ones are a standing problem rather than a theoretical one.

The checks worth running once a year:

• Confirm no camera still has a default or shared password. If every camera on site uses the same one from the install, that is one credential away from all of them.
• Update firmware where the manufacturer has published fixes. Read the release notes first and do the recorder before the cameras where the vendor says so.
• Confirm nothing is exposed straight to the internet through port forwarding. Remote access should go through the manufacturer’s relay service or a VPN.
• Check time sync. Cameras drift, and a timestamp two hours out makes footage hard to line up with a door log or a till receipt.
• Identify cameras the manufacturer no longer supports. An end-of-life camera getting no security fixes is a replacement decision, not a maintenance one.

Disk health, which has a predictable ending

Surveillance recorders write continuously, which is a heavier duty cycle than an office computer’s drive ever sees. Drives in them wear out, and the failure is not usually sudden.

Most recorders expose drive health, often under a SMART status page. Check it, and check the recorder’s own log for read errors and reallocated sectors, which show up long before a drive dies. Also check the physical box: recorders sit in dusty rooms, the fan clogs, and heat shortens drive life further.

Two practical points. Recorders should be on a small uninterruptible power supply, because unclean shutdowns are hard on drives and on the file system holding your footage. And when a drive is replaced in a multi-drive recorder, the rebuild has to finish before the system is protected again, so confirm it completed rather than assuming.

The annual check, in order

1. Open the live grid and count. Every camera that should be there, present and giving a picture.
2. Open playback and confirm each camera has footage from last night, not just a live image.
3. Record the oldest available date per camera. Compare against the retention you intended.
4. Walk the site with the app open and look at each camera physically: aim, focus, webs, housing condition, cable at the entry point.
5. Check the recorder: disk health, temperature, fan noise, error log, free space.
6. Check time sync across all cameras against one reference.
7. Check firmware versions and passwords.
8. Test an export. Pull a two minute clip to a USB stick and open it on a different computer.
9. Switch on video loss and disk failure notifications if they are off, and point them at an address somebody reads.
10. Write down what failed. That list is next year’s work order.

What this means for your building

A self-storage site in Courtice is where retention drift does the most damage. Dozens of cameras, constant motion from vehicles and doors, and a shared disk pool. Add four cameras to a new building on the site and retention across the whole system quietly drops, so the incident from five weeks ago is gone even though the setting still says 60 days.

A clinic in Oshawa fails differently. Few cameras, all indoors, nothing gets knocked or webbed, and the whole risk sits in one recorder in a cupboard with no ventilation and a drive that has been writing continuously for six years.

A yard in Brooklin fails physically: webs on the infrared, a camera bumped out of aim by a bin, and one drop that took water at the connector.

If you do nothing else

Check that every camera has footage from last night, and check the oldest date on the disk. Those two take ten minutes on a phone and catch the majority of what goes wrong.

Everything else on the list is worth doing yearly, and it is the whole content of a CCTV maintenance plan if you would rather it was somebody else’s job to remember.