A test server for sampling how long it takes mobile & web clients to make various types of requests to a server doing common request patterns. The idea is you can set this server up in various locations (think DCs) around the globe or country and test what the latency and experience for users will be as you vary the size and style of requests. You can also use the server to see how different CDNs might perform when sitting in front of your server and caching content.
The basic flow is as follows:
- Client gets a collection of URLs to call from server, a mix of GET & POST calls.
- Client executes the URLs timing each one.
- Client reports back to server all the timing data for each call and metadata about the client's connection.
####Clone the repo & build
./gradlew clean shadowJar
####Setup the Cassandra DB
Options:
- run standalone
- run ccm
- point at your internal cluster
Execute the scheme file in db/scheme.cql on your setup.
####Configuration
The server takes the following config
{
"port" : <port>,
"default_test_base_url" : "<default http://localhost>",
"data_retention_seconds" : <default 86400>,
"cassandra": {
"seeds": ["localhost"],
"reconnect": {
"name": "exponential",
"base_delay": 1000,
"max_delay": 10000
}
}
}
For example:
{
"port" : 7345,
"default_test_base_url" : "http://localhost",
"data_retention_seconds" : 86400,
"cassandra": {
"seeds": ["localhost"],
"reconnect": {
"name": "exponential",
"base_delay": 1000,
"max_delay": 10000
}
}
}
Field breakdown:
port
the port the server listens ondefault_test_base_url
the base url used in creating the bootstrap sample datadata_retention_seconds
the seconds to allow the upload test data from users to live in the DB, this defaults to a day.cassandra
The config to initialize the Cassandra driver with, see vertx-cassandra for more details.
####Running Server
#start the server
java -jar build/libs/ChronoServer-0.5.0-fat.jar -conf conf.json -instances 2
Gets a set of URLs for the client to execute.
Parameters:
test_batch
- a named collection of URLs you want to pass back to a client to run. Defaults to the bootstrap collection if not specified.
Response:
####/api/v1/report - POST
Allows a user to upload detailed information on the timings for all the calls it made.
Parameters:
####/api/v1/timing - GET Sub APIs
- /static - grabs
N
data from memory - /static_cached - grabs
N
data from memory and puts the appropriate cache headers on - /dynamic - grabs
N
data from the DB - /dynamic_cached - grabs
N
data from the DB and puts the appropriate cache headers on
Required Common Parameters:
unit
- the unit of the data to get, eitherkb
ormb
size
- the size of data to get back
Response:
{
"data" : "<a bunch of random data matching your parameters>"
}
####/api/v1/timing - POST Sub APIs
- /store - upload
N
amount data to the DB, server returns after it is successfully written to Cassandra - /store_mem - upload
N
amount data to server memory, server returns after it gets the full body of the post
Parameters:
####/healthcheck - GET
Standard health check for things like a LB to hit.
####/api/v1/timing_streaming - SockJS interface
See: streaming namespace and sample client below.
Piping curl output to something like jsonpp makes it much more readable.
###curl requests
#start the server
java -jar build/libs/ChronoServer-0.5.0-fat.jar -conf conf.json -instances 2
#grab a sample config package, take note of the api_key
curl -v "http://localhost:7345/api/v1/test_endpoints" | jsonpp
#grab static data with no caching
curl -v --header "X-API-Key: [api_key here]" "http://localhost:7345/api/v1/timing/static?unit=kb&size=1" | jsonpp
#grab static data with caching
curl -v --header "X-API-Key: [api_key here]" "http://localhost:7345/api/v1/timing/static_cached?unit=kb&size=1" | jsonpp
#grab data from the DB with no caching
curl -v --header "X-API-Key: [api_key here]" "http://localhost:7345/api/v1/timing/dynamic?unit=kb&size=1" | jsonpp
#grab data from the DB with caching
curl -v --header "X-API-Key: [api_key here]" "http://localhost:7345/api/v1/timing/dynamic_cached?unit=kb&size=1" | jsonpp
For testing POST calls via the store api Postman is great for this.
###sample SockJS client
<html>
<script src="http:////cdn.jsdelivr.net/sockjs/0.3.4/sockjs.min.js"></script>
<script src="vertxbus-3.0.0.js"></script>
<!-- You're going to want to turn off security in your browser when running this probably -->
<script>
// this is acquired from the server
var apiKey = ""
var sock = new SockJS('http://localhost:7345/api/v1/timing_streaming');
// download via socket
var getData = function(address, unit, size) {
sock.send(JSON.stringify({address: address, "api_key": apiKey,
body : {unit : unit, size: size}}));
};
// helper func for generating fake data
var getKb = function(number) {
var data = ['a', 'b', 'c', 'd', 'e', '1', '2', '3', '4', '5'];
var str = "";
var totalKb = 1024 * number;
for(var i = 0; i < totalKb; i++) {
str += data[Math.floor((Math.random() * 10))];
}
return str;
};
// upload via socket
var putData = function(size) {
sock.send(JSON.stringify({address: "streaming.store", "api_key": apiKey,
body : {test_batch: "some_batch", unit : "kb", size: size, data: getKb(size)}}));
};
// get the config, ie api key
var getApiKey = function() {
sock.send(JSON.stringify({address: "streaming.config", "api_key": "", body : {}}));
}
sock.onopen = function() {
console.log('connected to server');
getApiKey();
};
sock.onmessage = function(e) {
console.log('message', e.data);
var data = JSON.parse(e.data);
// once we have the api key make some test calls
if(data.type == "streaming.config") {
apiKey = data.api_key;
getData('streaming.static', "kb", 1);
getData('streaming.dynamic', "kb", 5);
putData(5);
}
};
sock.onclose = function() {
console.log('lost connection to server');
};
</script>
<body>
</body>
</html>