Replica Set Latency Monitoring Block for Solid

Edit MEMBERS for the nodes, READS and WRITES for the p99 series (one array of 60 minute samples per node, primary first) and LAG_ONE and LAG_TWO for each secondary's lag in tenths of a second. ALERT and LAG_LIMIT set the two lines. The header sentence, the badge values and the stale-read sentence are all derived from these arrays.

A secondary applies the primary's changes after a delay, its replication lag. A read sent to that secondary can therefore return data up to that old. The sentence names the secondary with the larger lag at the cursor's minute and gives the figure, so you can see how long a lagging node was a risk for reads that accept stale data.

It is a native range laid over the three plots, named Time, whose value text is the minute. Arrow keys move it one minute, Home and End jump to the ends, and dragging scrubs it. It starts on the minute of the worst lag. Switching between Reads and Writes keeps the minute. A polite status line speaks the reading about 600 ms after the last change.

All of it is typed: the series and the lag figures are examples, and no database is contacted. The product is made up and no vendor is named in the panel. In your app feed the same shapes from your database's metrics.

Yes. Every line is the same foreground stroke, told apart by its row and its label; the alert line is a thin rule, and the cursor is a hairline with a circle on each line. The figures beside each row and the lag badges give every value in words and numbers.

The node labels narrow, the time labels thin to every other one, and each lag row puts its bar under the name and badge. The two tabs and the rows keep their heights when you switch.