Agent connectivity: SIP/WS and connection errors

1. Description

A dashboard for network diagnostics of agent connectivity. It shows whether an agent is really reachable (SIP registration, WebSocket), whether the system moved them to offline because the connection was lost, and which attempts to connect a call to them failed. Answers the question "the agent is online, but calls and chats are not routed to them — why".

Refreshes automatically every 10 seconds.

Consists of:

Filters: the "Teams", "Agents", "Queues", "Alert window, min" and "Row limit" drop-down lists;
Current state stat block (6 stats);
The "Agents: current SIP/WS status" table;
The "Confirmed system disconnects by minute" chart;
The "Failed agent legs by cause" chart;
The "System auto-offline history (WS/SIP)" table;
The "Failed agent connections — diagnostic layer" table.

1.1. Drop-down lists

Filter	Purpose
Teams	Selects the teams to display data for
Agents	Selects agents
Queues	Selects queues. Affects the failed-connection panels
Alert window, min	The interval used by the "Disconnects in alert window" and "Agent-leg errors in alert window" stats: 5, 15, 30 or 60 minutes. Does not affect any other panel
Row limit	The maximum number of rows displayed in the dashboard tables

1.2. Current state stat block

Stat	What it shows	Colours
Online / break_out	The number of agents in `online` or `break_out` status — logged in, regardless of whether they are handling interactions right now. Current state; the Grafana time range does not affect this stat	Green if at least one agent is found; `0` is blue
Online without WS	Of the `online` / `break_out` agents — how many currently have no active WebSocket connection. Agents with an active	`0` — green; `1` or more — red
Online without SIP	Of the `online` / `break_out` agents — how many currently have no active SIP connection. Agents with an active	`0` — green; `1` or more — red
Current auto-offline	How many agents the system is currently keeping offline because WS or SIP is missing (causes `system/ws`, `system/sip`, `system/ws/sip`). These are current unresolved cases, not history over a period	`0` — green; `1` or more — red
Disconnects in alert window	The number of confirmed auto-offline events within the interval set by the "Alert window, min" filter. Independent of the global Grafana time range	`0` — green; `1` or more — red
Agent-leg errors in alert window	The number of failed attempts to connect a call to an agent within the same interval. Counts `USER_NOT_REGISTERED`, `NO_USER_RESPONSE`, `NO_ANSWER`, `LOSE_RACE`, route/transport errors, `GATEWAY_DOWN` and `NORMAL_TEMPORARY_FAILURE`	`0` — green; `1` or more — orange. Orange rather than red because not every failed attempt confirms a network fault

1.3. The "Agents: current SIP/WS status" table

Consists of the following columns:

Column	What it means
Agent / Agent ID	The agent's name and internal identifier
Team	The team the agent belongs to
Status	The computed status: `online`, `offline`, `break_out`, etc.
Call channel	The state of the voice channel: `waiting`, `bridged`, `processing`, `wrap_time`, etc.
WS	WebSocket connection state: `CONNECTED` / `DISCONNECTED`
SIP	SIP registration state: `CONNECTED` / `DISCONNECTED`
Push	`ACTIVE` if the agent has a valid Webitel mobile Push session, otherwise blank. It only means the app can be woken by a push: it does not confirm a SIP or WS connection and does not guarantee delivery of any specific message
Diagnosis	The overall assessment of the connection: `OK`, `WARNING` or `CRITICAL` with a reason (for example, `CRITICAL: auto-offline`)
Status since / Channel since	The time of the last status change and, separately, of the last call-channel change
Last WS / Last SIP	The time of the last activity over WebSocket and over SIP
Status reason	The technical payload (for example, `system/ws/sip` — the status was set by the system because WS and SIP dropped at the same time)

Rows are sorted CRITICAL first, then WARNING, then OK; within a group — by team and agent name.

The table shows the current state, so the global Grafana time range ("Today", "Last 6 hours") does not filter it. A past date in the "Last WS" and "Last SIP" columns means there has been no activity over that connection since then — it is not a display error.

1.4. The "Confirmed system disconnects by minute" chart

A bar chart grouped by minute. The WS, SIP and WS + SIP series are stacked on top of each other; the legend below shows the total number of events of each type for the period.

Depends on the global Grafana time range. No data means there were no confirmed system disconnects in the selected period.

1.5. The "Failed agent legs by cause" chart

A bar chart of failed attempts to connect a call to an agent, grouped by minute and by termination cause. Each cause has its own colour: USER_NOT_REGISTERED, NO_ANSWER, NO_USER_RESPONSE, LOSE_RACE, route/transport causes and others.

Depends on the global Grafana time range and on the "Queues" filter. No data means there were no failed attempts in the selected period.

1.6. The "System auto-offline history (WS/SIP)" table

The history of confirmed system disconnects, one row per episode. Depends on the global Grafana time range.

Consists of the following columns:

Column	What it means
Start / End	The time the episode started and ended
Duration, sec	The length of the episode. For `ONGOING` it is counted from the start of the offline period to the present moment; for `ENDED` — until the connection was restored or the status changed again. Despite the "sec" label, the value is displayed in convenient units: seconds, minutes, hours or weeks
Agent / Agent ID	The agent the episode concerns
Team	The agent's team
Disconnect	What exactly dropped: `WS`, `SIP` or `WS + SIP`
Payload	The technical cause tag
State	`ONGOING` — the episode is still in progress; `ENDED` — it has finished

1.7. The "Failed agent connections — diagnostic layer" table

A detailed list of failed attempts to connect a call to an agent (agent leg), one row per attempt. This is the section to open once the "Agent-leg errors in alert window" stat has been triggered.

Depends on the global Grafana time range and on the "Queues" filter. No data means there were no such attempts for the selected period and queue.

Consists of the following columns:

Column	What it means
Time / Agent / Team / Queue	When the attempt was made, to whom, and from which queue
Cause	The agent leg termination cause code
SIP code	The SIP response code
Attempt ID / Agent Call ID	The identifiers of the distribution attempt and of the call on the agent's side
Diagnostics	The interpretation of the cause (see below)

Values of the "Diagnostics" column:

Value	What it means
`CONFIRMED: SIP not registered`	The agent's SIP was not registered at the time of the attempt
`LIKELY: route/transport failure`	A probable routing or transport problem
`POSSIBLE: endpoint did not answer`	The agent's device did not respond
`INDIRECT: race/bridge failure`	An indirect sign that does not in itself prove a network fault. This is where `LOSE_RACE` belongs
`OTHER`	Another cause

Only the CONFIRMED value unambiguously confirms a problem with the agent's connection. The rest require verification.