Watch your server: a health report from a real shift

A watch samples your server (or your whole cluster) for 5 minutes to 2 hours while your agents work, then tells you what went wrong, when, and what to do about it. It is read-only: it changes nothing on the server and never stops a call.

Starting a watch

1. Open your server, or your cluster, and find the "Watch this server" card.
2. Pick how long: 5, 15, 30 or 60 minutes, or 2 hours. 15 to 30 minutes of a busy shift is usually enough.
3. Click Start watch. If nobody is signed in and no calls are running, you are warned first: a watch of an idle floor can only report that the floor was idle. You can still start it.
4. Leave it running. You can stop it early from the same card; the report then covers what ran.

Only the account owner or an admin member can start or stop a watch. Anyone with access to the server can read the reports. One watch runs at a time per server or cluster.

What the report shows

• A summary in plain English: was the server healthy, and if not, what mattered most.
• Compared with the last watch: the key numbers next to the last watch of the same server that had agents working, with counts scaled to an hour so a 15-minute watch reads fairly against a 60-minute one. Anything clearly worse is marked.
• Findings: each problem with the exact times and numbers, and what to do. Database lock queues, dispositions that waited, agent screens that lagged or logged out, your office internet dropping, carriers refusing calls, campaigns dropping calls, rooms or connections running out, and errors piling up in the logs.
• Charts for every server: agents, database lock queues, agent-screen response, web traffic by network, load, and calls on each dialer, with the problem moments shaded.
• Calls by carrier: calls dialled, answered, busy or unanswered, refused by the carrier, the SIP answers it gave, and ring time. VICIdial does not record post-dial delay, so ring time (from dialling to answer or give-up) is shown instead.
• Campaigns: calls dialled, answered and dropped during the watch, with the drop rate. The usual limit is 3% of answered calls.
• Answering-machine detection: how many calls each verdict got, and how long detection took.
• Agents: one row per agent with calls, talk and waiting time, time on calls, ready, paused and in disposition, lagged pauses, logouts and how fast their screen answered. Agents with problems come first.
• Log lines: what Asterisk, the web server and the database wrote during the watch, grouped by message with real examples.

Email and PDF

When the watch ends, the report is emailed to whoever started it, with the main charts in the email and the full report attached as a PDF. The report page has the same PDF to download. Reports are kept for 30 days.

Privacy

The watch never copies call recordings or lead details. In the log examples, phone numbers are cut to their first digits, passwords and keys are replaced with ***, and addresses are shortened to their first two parts. Web traffic is grouped by network, never by full address.

If the report says the floor was idle

Nothing was wrong, but nothing was tested either. Run the watch again while agents are signed in and calls are flowing.