Tweaked README, changed default log collection behavior on VerifiableProducer

tests/README.md

@@ -15,7 +15,7 @@ This quickstart will help you run the Kafka system tests on your local machine.
 * Install Vagrant Plugins:
 
         # Required
-        $ vagrant plugin install vagrant-hostmanager
+        $ vagrant plugin install vagrant-hostmanager vagrant-cachier
 
 * Build a specific branch of Kafka
        
@@ -57,12 +57,15 @@ installing dependencies and updates on every vm.):
         
 EC2 Quickstart
 --------------
-This quickstart will help you run the Kafka system tests using Amazon EC2. As a convention, we'll use "kafkatest" in most names, but you can use whatever you want. 
+This quickstart will help you run the Kafka system tests on EC2. In this setup, all logic is run
+on EC2 and none on your local machine. 
 
 There are a lot of steps here, but the basic goals are to create one distinguished EC2 instance that
 will be our "test driver", and to set up the security groups and iam role so that the test driver
 can create, destroy, and run ssh commands on any number of "workers".
 
+As a convention, we'll use "kafkatest" in most names, but you can use whatever name you want. 
+
 Preparation
 -----------
 In these steps, we will create an IAM role which has permission to create and destroy EC2 instances, 

tests/kafkatest/services/verifiable_producer.py

@@ -23,7 +23,7 @@ class VerifiableProducer(BackgroundThreadService):
     logs = {
         "producer_log": {
             "path": "/mnt/producer.log",
-            "collect_default": True}
+            "collect_default": False}
     }
 
     def __init__(self, context, num_nodes, kafka, topic, max_messages=-1, throughput=100000):
@@ -47,8 +47,6 @@ class VerifiableProducer(BackgroundThreadService):
             data = self.try_parse_json(line)
             if data is not None:
 
-                self.logger.debug("VerifiableProducer: " + str(data))
-
                 with self.lock:
                     if data["name"] == "producer_send_error":
                         data["node"] = idx

tests/kafkatest/tests/replication_test.py

@@ -97,7 +97,12 @@ class ReplicationTest(Test):
         self.logger.info("num consumed:  %d" % len(self.consumed))
 
         # Check produced vs consumed
-        self.validate()
+        success, msg = self.validate()
+
+        if not success:
+            self.mark_for_collect(self.producer)
+
+        assert success, msg
 
     def clean_shutdown(self):
         """Discover leader node for our topic and shut it down cleanly."""
@@ -142,7 +147,7 @@ class ReplicationTest(Test):
             # Collect all the data logs if there was a failure
             self.mark_for_collect(self.kafka)
 
-        assert success, msg
+        return success, msg
 
     def test_clean_shutdown(self):
         self.run_with_failure(self.clean_shutdown)