Diagnosing intermittent agent disconnects across a shift

Intermittent disconnects are the hardest kind to fix because they never happen while you're watching. One agent drops at 10am, another at 2pm, and by the time you log in everything looks fine. The trick is to stop trying to catch it live and instead read the history three ways: the Agent LAGGED Report, the Agent Latency Report, and the Latency Gaps Report. Together they turn "it just disconnects sometimes" into a pattern you can act on.

Three reports, three angles

Each report shows a different face of the same problem:

• The Agent LAGGED Report flags the moments an agent's screen lagged behind the server. It records the agent log ID, user ID, server IP, event time, lead ID, campaign, status, dial method, and the Asterisk uniqueid — enough to pin a lag to one agent on one Campaign at one instant.
• The Agent Latency Report shows the agent screen's web connection Latency over the course of the day, with a chart per user, so you can see a connection getting slow before it fails.
• The Latency Gaps Report shows the windows where the screen stopped reporting in entirely — the gaps that line up with an actual drop and a torn-down Agent session.

Reading them in order

Start with the LAGGED Report to find when and to whom it happened. Take those timestamps to the Agent Latency Report to see whether the connection was already degrading. Then check the Latency Gaps Report to confirm the screen actually went silent at that moment. When all three agree on a time and an agent, you have a real event rather than a complaint — and the Status (lead status) recorded at that moment tells you what the agent was doing when it dropped.

flowchart TD
  A[Agent reports a random drop] --> B[Agent LAGGED Report]
  B --> C[Find the agent and event time]
  C --> D[Agent Latency Report]
  D --> E{Connection slow before drop?}
  E -->|Yes| F[Latency Gaps Report confirms silent window]
  E -->|No| F
  F --> G{Pattern across agents?}
  G -->|One agent| H[Network or workstation]
  G -->|Many at once| I[Shared uplink or DB server load]

What the pattern tells you

Disconnects almost always come down to one of three culprits: the agent's network, the agent's workstation, or load on the database server behind the dialer. The shape of the data tells you which:

• Lag and gaps tied to a single agent across many days point at that agent's connection or PC — a flaky Wi-Fi link, a tired laptop, a browser drowning in tabs.
• Lag and gaps that hit many agents at the same clock time point at a shared cause — a saturated office uplink or a database server straining under report queries or a heavy lead load.

For more on the lag events themselves, read why agents get logged out as LAGGED, and keep the VICIdial troubleshooting playbook open while you work through it.

When the cause is database-server load, no amount of agent-side tuning helps — you need headroom on the box. VICIfast runs your dialer on a dedicated, hardened server that's live in under 40 seconds, so an overloaded shared host stops being the thing dropping your agents. See our plans and pricing.