I'm trying to do some load testing on my Play 2.0.2 app, but I keep running into the following exception:
Cannot invoke the action, eventually got an error: Thrown(akka.pattern.AskTimeoutException: Timed out)
I'm using http://blitz.io/ to do my load testing.
My tech stack is Play + mysql.
I'm not using any promises in my app.
Here's the customized akka configuration:
#default timeout for promises
promise.akka.actor.typed.timeout=10s
play {
akka {
event-handlers = ["akka.event.slf4j.Slf4jEventHandler"]
loglevel = ERROR
actor {
deployment {
/actions {
router = round-robin
nr-of-instances = 100
}
/promises {
router = round-robin
nr-of-instances = 100
}
}
retrieveBodyParserTimeout = 10 second
actions-dispatcher = {
fork-join-executor {
parallelism-factor = 100
parallelism-max = 200
}
}
promises-dispatcher = {
fork-join-executor {
parallelism-factor = 100
parallelism-max = 200
}
}
websockets-dispatcher = {
fork-join-executor {
parallelism-factor = 100
parallelism-max = 200
}
}
default-dispatcher = {
fork-join-executor {
parallelism-factor = 100
parallelism-max = 200
}
}
}
}
}
One more thing: After the application runs for a day or so, it starts slowing down. Upon checking the resources (using htop), I see that the cpu utilization is 100%, so its clear that the node is stuck doing too much work that it's response gets delayed immensely. Furthermore, I can see that a lot of java instances have been spawned (8-10) on average. Why is that? Is it natural for a Play app to spawn so many java instances? About mem usage, the java play instance gets 512MB-1536MB at the time of spawning and I can see from the stats that the memory is about 50-70% used on average so probably thats not where the problem is (possible mem leaks etc).
I'm running on an extremely tight schedule, so I would appreciate any help/pointers I can get into investigating (or possibly fixing) this issue.
Thanks a bunch!
i think you might have some blocking IO, which is why you need the
retrieveBodyParserTimeout = 10 second
hack. Try making it async.