lambda container reuse state issue

handler.js

/* 
 The current container reuse strategy in AWS Lambda makes it very hard to ensure consistent environments between invocations. 
 Managing environments is currently the responsibility of the user. There should be some additional cleanup tasks ran after 
 an invocation to ensure at least a freah new nodejs context. As an example I added the setInterval issue we've seen.
*/

// new container a
// * invocation 1 on container a
//   - using setInterval to check remaining time every 1 sec
//   - the handler will call context.success() after 45 sec without clearing setInterval
// * invocation 2 on container a
//   - if the container is reused the setInterval will fire immediately on invocation and call context.fail prematurely

exports.handler = function (event, context) {
  setInterval(function () {
    // around 55 seconds if we haven't finished cleanup and exit
    if (context.getRemainingTimeInMillis() < 5000) {
        // we do /tmp cleanup and child process cleanup here.
        // i.e. rm -rf /tmp/pg-import & killall gawk
        
        // invocation a2 exits prematurely here
        context.fail('we hit timeout');
    } 
  }, 1000);
  
  setTimeout(function () {
    // invocation a1 exits here
    context.success();
  }, 45000);
};