VICIfast
ENES
Guides & tutorials

Reports and tools that slow a live VICIdial floor

Which reports, stats tools and background jobs make a VICIdial floor stall mid-shift, what each one cost on floors we measured, and how to keep them out of the way.

VICIfast Support
··4 min read
Analytics dashboard on a monitor — illustration for: Reports and tools that slow a live VICIdial floor
Photo by RDNE Stock project on Pexels

The floor was fine at nine. By eleven, statuses hang after the agent clicks, screens stutter, and calls drop while agents sit ready. Nobody touched the dialers. What changed is usually something somebody opened: a report, a wallboard, a stats page, an outside tool.

Most reports are harmless. The ones that hurt share one habit: they read VICIdial's call log for a long time while the floor is busy writing to it.

Why a read can stop a floor

VICIdial keeps its call records in MyISAM tables. A MyISAM table locks as a whole while a read runs on it, so every write that arrives in the meantime waits for the read to finish.

On a live floor the writes never stop. Each Disposition, the status an agent picks when a call ends, is a write. So is handing an answered call to a free agent, which VICIdial does through the database. When that hand-off waits, the drop timer keeps counting, and the call can end up in your Drop rate: dropped calls as a share of calls a person answered.

In a cluster all of this happens on the database and web server, the one machine every dialer depends on. Our guide to VICIdial clusters shows where it sits.

flowchart LR
  A[Day-long report] --> D[(Call log)]
  B[Stats page or wallboard] --> D
  C[Outside tool or stock job] --> D
  D --> E[Table locked while it reads]
  E --> F[Status saves wait]
  E --> G[Call hand-offs wait]
  G --> H[Calls drop with agents ready]

What we measured, tool by tool

These come from live floors we work on. Each figure is from one or two floors, so read them as what can happen, not what will.

What ran during the shiftWhat it did
A day-long countHeld up to 334 statements for as long as 4 seconds (60-agent floor)
A stats tool reading the call log58 lock waits a second; 13 after it was moved off the call log
VICIdial's stock scan for unanswered calls8 to 27% of database time on two floors
A CRM polling the VICIdial API once per agent25 to 39% of all web requests

The stats tool is the one that surprises people. It only reads, so it looks innocent, but it reads all shift long. Moving that single tool off the call log cut lock waits by more than three quarters.

The scan for unanswered calls isn't something anyone installed. It's a stock VICIdial job that looks through the call log, and without an index it does far more reading than it needs to.

The CRM connector is a different kind of load. Asking about each agent separately piles requests onto the same machine that serves the agent screens. One request for all logged-in agents, cached and polled less often, gets the CRM the same answer.

Finding the one on your floor

Start with the shape of the stall. If one agent hangs and everyone else is fine, look at that agent's browser and connection. If the whole floor freezes at the same second and recovers together, something is holding the call tables.

Note the times. Watch the Real-time report, VICIdial's live view of agents and calls, and write down when agents sit ready while calls drop or screens stop moving.

Then check reports. The Admin Report Log Viewer lists every report pulled, who ran it and how long it took to compile, and it highlights the slow ones. Line those rows up against your times. Reading the Admin Report Log Viewer walks through the columns.

Tools are harder, because they don't show up in that log. List everything that reads your database during the shift: wallboards, stats pages, CRM connectors, scripts somebody wrote years ago. Each one is a suspect until you know what it reads and how often.

CPU and memory graphs usually look fine while this happens. A locked table is waiting, not working, so a bigger processor won't shorten the stall.

Keeping heavy reads out of the shift

  • Run day-long counts and exports before the floor starts or after it stops.
  • Point stats tools away from the live call log, or have them refresh far less often.
  • Give the stock unanswered-call scan an index; in our test that cut its cost by about 95%.
  • Ask your CRM to fetch all logged-in agents in one request instead of one per agent.
  • Keep the log tables small with VICIdial's log archiving, so the reads that do run finish sooner.

If one server still carries your floor, every fix here works on it too; if that's your size, start on one of our plans. On the clusters we run, we find what holds the call tables during a shift and move it. When a shift still goes wrong, we can record the next one second by second: database locks, each dialer's load, every call that waited.

Planning a floor big enough to need separate machines? Size your cluster and see today's prices.

About VICIfast

We run VICIdial servers for call centres, so you don’t have to. You get the dialer set up, secured and kept running — we handle the server, the updates and the backups. Bring your own phone carrier at no markup, or use VICIfast Voice, our own US calling and numbers, from $0.0085 a minute. From $49 a month per server.

Questions? Call +1 636 556 0022 and talk to someone who runs dialers.

Citing this article

VICIfast Engineering. “Reports and tools that slow a live VICIdial floor”. VICIfast LLC, November 21, 2026. Retrieved from https://vicifast.com/blog/vicidial-reports-that-slow-the-floor

Have questions?

Related posts

You might be interested in

Comments

Comments are reviewed before they appear. We never publish your email.

No comments yet - be the first.