How does Timer component polls? - apache-camel

Question

Let's say I have following time component:
from("timer://foo?period=1000").setBody(constant("select * from customer")).to("jdbc:testdb").to("beanRef:processResult");
How does timer component work here? Does it reads from database in every 1 sec or waits for bean to finish the processing?
If bean is still processing the earlier result and timer will keep polling the database then it will create a bottleneck. Is there any way to avoid it?

Answer 1

Okay, Update: looking at the source code, the timer component relies on the java TimerTask implementation. And your question is already answered here: Is Java's Timer task guarenteed not to run concurrently?
Short answer: one single thread executes the trigger and the routes connected to it, so there will be no concurrent execution.
That said, you might want to controll the execution a bit. It is recommended with Timer Tasks (and hence Camel timers) to have a margin between the period in the timer and the max task execution time.
You can use a SEDA component (with concurrentConsumers=[num threads]) in between to fine grain controll the execution with a work queue. The timer will finish it's task right away while the real route can continue to process.
from("timer://foo?period=1000")
.to("seda:startRoute");
from("seda:startRoute")
.setBody(constant("select * from customer"))
.to("jdbc:testdb").to("beanRef:processResult");
Each event will stack up non the less, so over time, you might want tune the route so that the period > avg route exec time.
You could add a shared boolean variable either in a singleton bean or in a static class:
public static synchronized boolean isRunning(){
return running;
}
public static synchronized void setRunning(boolean isRunning){
running = isRunning;
}
The variable should telling weather the route is running or not and filter timer events that occurs while the variable is true. Just hook up a few processors/bean-calls to handle this.