如何为Gatling中asLongAs循环的重试请求设置超时？

Great question! When using Gatling's asLongAs for polling a service until you get a result, you don’t have to rely on infinite loops—there are two clean ways to add a timeout. Let’s adapt your code with both approaches:

Option 1: Use asLongAs's Built-in maxDuration Parameter #

Gatling’s asLongAs has an overloaded method that accepts a maxDuration parameter, which is the easiest way to set a hard timeout for your polling loop. You’ll need to import Scala’s duration utilities to use readable time units like seconds.

Modified Code #

First, add the import at the top of your file:

import scala.concurrent.duration._

Then update your scenario:

scenario("my-scenario")
 .feed(feeder)
 .exec(
  http("post-to-A")
  .post("/execution")
  .body(StringBody(
   """{
    | "information": ${INFORMATION}
    |}""".stripMargin
  ))
  .asJSON
  .check(status.is(200))
 )
 .exec(_.set("result", ""))
 // Set a 60-second timeout; loop stops either when result is non-empty OR 60s passes
 .asLongAs(session => session("result").validate[String].get != "", maxDuration = 60 seconds) {
  exec(
   http("get-to-B")
   .get("/result")
   .check(status.is(200))
   .check(jsonPath("$.result").saveAs("result"))
  )
  // Add a pause to avoid overwhelming Service B with too-frequent requests
  .pause(1 second)
 }
 // Optional: Handle the timeout case after the loop ends
 .exec(session => {
  val result = session("result").validate[String].get
  if (result.isEmpty) {
    // Log or mark the session as timed out for reporting
    session.set("polling_timed_out", true)
  }
  session
 })

How It Works #

• The maxDuration = 60 seconds tells Gatling to terminate the loop after 60 seconds, regardless of whether the result is still empty.
• Adding a pause is crucial to prevent your test from flooding Service B with requests—adjust the duration based on how quickly you expect the result to be ready.

Option 2: Manually Track Poll Start Time (For Custom Logic) #

If you need more control (e.g., combining timeout with max retry count), you can track the start time of your polling in the session and add a time check to your loop condition.

Modified Code #

scenario("my-scenario")
 .feed(feeder)
 .exec(
  http("post-to-A")
  .post("/execution")
  .body(StringBody(
   """{
    | "information": ${INFORMATION}
    |}""".stripMargin
  ))
  .asJSON
  .check(status.is(200))
 )
 // Save the current timestamp when polling starts
 .exec(session => session.set("poll_start_time", System.currentTimeMillis()))
 .exec(_.set("result", ""))
 // Loop only if result is empty AND we haven't exceeded 60 seconds
 .asLongAs(session => {
  val result = session("result").validate[String].get
  val startTime = session("poll_start_time").validate[Long].get
  val elapsedTime = System.currentTimeMillis() - startTime
  result.isEmpty && elapsedTime < 60000 // 60000ms = 60 seconds
 }) {
  exec(
   http("get-to-B")
   .get("/result")
   .check(status.is(200))
   .check(jsonPath("$.result").saveAs("result"))
  )
  .pause(1 second)
 }

How It Works #

• We store the start time of the polling phase in the session.
• The loop condition now checks two things: whether the result is still empty, and whether less than 60 seconds have passed since polling started.
• This approach lets you add extra logic (like counting retries) alongside the timeout check if needed.

Key Notes #

• Adjust Timings: Tweak the maxDuration or elapsed time threshold to match your service’s expected execution time.
• Reporting: Use the optional post-loop exec block to mark timed-out sessions, which you can then analyze in Gatling’s test reports.
• Avoid Over-Polling: Always include a pause between polling requests to be kind to your target service.

内容的提问来源于stack exchange，提问作者riccardo.cardin