Monitoring
Overview dashboard
Read fleet health, throughput, lag and the pipelines that need attention on the Overview page.
The Overview page is the first page you see after you sign in. It shows whether your pipelines are healthy, how much data is moving, and which pipelines need you. It covers the account you are working in; other accounts have their own Overview.
Check that the page is live
The first line of the page tells you how fresh the numbers are:
| You see | Meaning |
|---|---|
Live · streaming from N pipelines · updated just now | The numbers are current. |
Live · … · updated 8s ago | The last update is a few seconds old. |
Stream · … · Reconnecting: data may be stale | No update for 15 seconds, or the connection to the control plane dropped. The browser reconnects by itself. |
| Connecting · … | The first update hasn't arrived yet. |
If the control plane can't be reached, the page shows an error panel such as Couldn't reach the control plane with a Retry button. It never shows zeros in place of an outage.
Fleet throughput
The two large figures are the fleet's events per second and data rate, summed over every running pipeline. Paused and dry-run pipelines add nothing.
- events / sec is the average over the last 60 seconds, so a pipeline that applies changes in bursts doesn't read as
0between bursts. - The caption reads
<volume> moved · last 24h. Once agents have reported long enough, it addshealthy <P>% of the time observed: the share of the last 24 hours in which your pipelines reported healthy. Loading and paused time counts against it.
Error rate and backlog
Two tiles sit beside the throughput figures:
| Tile | What it shows |
|---|---|
| Error rate | The percentage of events that failed to apply, across running pipelines. It is red above 0, because a failed apply means the target no longer matches the source. |
| Backlog | Records captured and not yet applied, for the pipeline with the largest backlog. A steady backlog is normal; a backlog that keeps growing is the signal. |
Fleet health
The Fleet health block shows the percentage of running pipelines that are healthy, and a caption such as 3 of 4 running healthy. Paused pipelines are left out, because pausing is deliberate. A pipeline whose agent is offline counts as running and not healthy. When nothing is running, the block shows –.
The legend and the bar below it use one color per status:
| Status | Color | Meaning |
|---|---|---|
| Healthy | Green | Streaming normally. |
| Loading | Blue | An initial load or reload is running, or automatic restart is bringing the pipeline back (shown as Restarting on the pipeline). |
| Degraded | Amber | Running with elevated lag or errors. |
| Failing | Red | Not making progress. The reason appears in Action required and on the pipeline. |
| Agent offline | Amber, lighter | The pipeline's agent hasn't reported for 60 seconds, so the pipeline isn't replicating and its lag and throughput are unknown. See Agents and brokers. |
| Paused | Gray | Stopped by an operator. |
| Dry run | Gray | A simulated run that moves no data. Counted with paused. |
Action required
Action required is your work queue. It lists up to four items, red before amber, and +N more require attention → when there are more. Each row opens the pipeline it names. When the queue is empty, it says No operator action required.
| Item | Color | Raised when |
|---|---|---|
| Add your first agent | Red | No agent is enrolled. Nothing replicates until one is online. |
| Create your first pipeline | Amber | An agent is enrolled and there are no pipelines. |
| Pipeline failing | Red | A pipeline is failing. The row shows the start of its failure reason. |
| Agent offline | Red | A running pipeline's agent is offline. For a pipeline with relay agents, either agent counts. |
| Lag above goal | Amber | The pipeline's lag stayed at or above its amber threshold for about the last minute. A single slow sample never raises it. |
| High error rate | Amber | A running pipeline's error rate is above 0. |
| Target paused | Amber | A target of the pipeline is paused, so the source spool grows until you resume it. |
| Assign an agent | Amber | A pipeline has no agent. |
| Add a target | Amber | A pipeline is a source monitor with no target. |
Pipelines
The Pipelines table lists every pipeline in two tabs: Active and Inactive (paused). View all → opens the full pipeline list.
Rows are sorted worst first: failing, agent offline, degraded, loading, paused, then healthy, and by lag within a status. Select a row to open the pipeline.
| Column | What it shows |
|---|---|
| Status dot and Pipeline | The status color, the pipeline name and its route, such as SQL Server · host/database → PostgreSQL · host/database. A pipeline with no target reads <source> source monitor. |
| Recent throughput | The average rate, a trend line and the current rate. For a failing pipeline or one whose agent is offline, the numbers show –. |
| Lag heatmap | One cell for each of the last 24 lag samples, oldest first, colored against the pipeline's own lag goal, so red always means trouble for that pipeline. An empty cell means no measurement. |
| Recent lag | A trend line and the current end-to-end lag, followed by the pipeline's lag goal, such as 3.2s • goal 10s. |
The lag figure shows – when the last measurement is 60 seconds old or older. A pipeline that has applied everything shows 0ms, and hovering says caught up.
The lag goal depends on the capture mode and the targets, and you can set your own. See Lag goals and colors.
Active incidents
Active incidents lists the three most recent open alerts. Select Ack to acknowledge an alert without leaving the page, or View all → to open Alerts.
Backfills in progress
Backfills in progress lists each running pipeline that still has a table to load, with the percentage loaded and a bar of tables migrated (green), in progress (blue) and still to do (gray). A paused pipeline leaves the list, because pausing cancels its load.
Your first visit
On an account with no pipelines and no alerts, the Overview shows Create your first pipeline with three steps (Connect, Select, Review) and a Create first pipeline button. You can add an agent during setup.
