Understanding duration monitors
msiebler 2700005RPQ Visits (9971)
A common misperception that we often see in support are the uses for duration monitors and for SLM 'latency' threshold types (like backend latency, internal latency and total latency).
Sometimes users want to have a mechanism to throw an alert when a single transaction has run very slowly and crossed some limit. Or they want to prevent any single transaction from running too slowly and they desire that a slow transaction would be stopped before crossing the limit.
The duration monitors and SLM are not designed to serve these functions. Rather, they are designed to report and take action when an average number of transactions has started to show some behavior. So for example, a good usecase for the duration monitor would be the following: say that a backend normally takes 10ms to respond. Now say that the backend is becoming overloaded and is slowing down, and it is now taking 100ms to respond. A duration monitor could detect this slowdown over a number of recent transactions and could send a notification to an administrator to investigate the issue. Note that again we are refering to the average latency over a number of transactions.