← All docs

Troubleshooting

Common problems, and where to look.

Where the logs are

A camera isn't recording

  1. Is it assigned to a node, and is that node online?
  2. Does the node have a storage path? The node list flags nodes without one.
  3. Is the camera enabled, and do its streams look right? Re-probe after any change on the camera itself.
  4. In Motion, Schedule, or Event mode, check whether footage is being discarded (see Recording).

Live view and playback stay on “connecting”

When a node streams directly to browsers (see Direct streaming), the browser connects to the node's client port, 4200 by default. If the browser's console shows a failed wss://<node>:<port>/live/… connection, check that the node's firewall allows that port: the installer creates an inbound rule named LarisVMS Node Client Endpoint. Installers before 0.215.1 could drop that rule when upgrading or taking over a node; run the current Node installer again, or add the rule by hand.

Active Directory sign-in or sync fails

Use Test connection under Settings → Active Directory; the last sync's error is shown on the same page.

Live view pauses in a background window

Chrome throttles windows that aren't visible, so a live wall on a second monitor can freeze and then catch up. Live view recovers on its own. To prevent it, add these to the Chrome shortcut's target:

--disable-backgrounding-occluded-windows --disable-background-timer-throttling

Every tile freezes, then reconnects at once

The node stalled, not the browser or the network. Look in the node log for Node process health: tick … ms late. A healthy node logs this about once a minute with lateness in single-digit milliseconds.

AI detection stops but recording carries on

Usually the GPU driver was reset. The Windows System event log shows nvlddmkm event 153 or Display event 4101 at that moment. The vision service restarts itself after 60 seconds without a result. Repeated resets point at the GPU, its driver, cooling, or power.

A node or proxy didn't come back after an update

Start its service (sc start LarisVMSNode, or LarisVMSProxy), then check updater-*.log for why the restart failed. A downloaded update that never got applied is also noted in the node's or proxy's own log when it next starts, and is offered again on a later check-in.

A node shows “clock”

The node's clock is more than a minute off from the server's. Fix its time sync. Clock skew affects schedules and the timeline.

A node keeps failing to check in

Use Reset auth from the node's ⋯ menu. It clears the node's check-in token, and the node re-syncs on its next heartbeat.

Also in the app under Help. View source on GitHub.