I am using httpcomponents to do http requests. I have set the connectionRequestTimeout, connectTimeout and socketTimeout to the same time(such as 8000ms). The system is in a high concurrency and most time it performs well, but some requests cost seconds which is the same as the timeout(~8000ms) per minutes. Here is the code snippet:
RequestConfig config = RequestConfig.custom().setConnectionRequestTimeout(TIMEOUT).setConnectTimeout(TIMEOUT).setSocketTimeout(TIMEOUT).build();
CloseableHttpClient client = HttpClientBuilder.create().setDefaultRequestConfig(config).build();
HttpUriRequest request = null;
switch (method) {
case GET:
String getUrl = url;
if (null != paramData) {
getUrl += "?" + paramData;
}
request = new HttpGet(getUrl);
break;
case POST:
...
default:
...
}
CloseableHttpResponse response = null;
try {
long start = System.currentTimeMillis();
response = client.execute(request);
long time = System.currentTimeMillis() - start;
// ***************************
// Sometime the log shows the cost is a few milliseconds more than TIMEOUT,
// but it does not throw any timeout exception and the response is fine.
// ***************************
LOG.debug("cost {}ms", time);
int resultCode = response.getStatusLine().getStatusCode();
HttpEntity entity = response.getEntity();
String resultJson = EntityUtils.toString(entity, UTF_8);
if (HttpStatus.SC_OK == resultCode) {
...
}
} catch (Exception e) {
...
} finally {
//abort the request
if (null != request && !request.isAborted()) {
request.abort();
}
//close the connection
HttpClientUtils.closeQuietly(client);
HttpClientUtils.closeQuietly(response);
}
The version of httpcomponents is 4.5.2 and the jdk is openjdk 1.8.0_92.
Moreover, should I use CloseableHttpClient as a singleton for better performance?
CloseableHttpClient should be used singleton for better performance.
You can also define the number of connection as per your requirement in HttpClientBuilder.
Also socketTimeout should be more that connectionTimeOut, This will give more time to read the socket.
Related
I am working on a legacy (Java 6/7) project that uses ProcessBuilder to request a UUID from the machine in an OS-agnostic way. I would like to use the Process.waitFor(long timeout, TimeUnit unit) method from Java 8, but this isn't implemented in Java 6. Instead, I can use waitFor(), which blocks until completion or an error.
I would like to avoid upgrading the version of Java used to 8 if possible as this necessitates a lot of other changes (migrating code away from removed internal APIs and upgrading a production Tomcat server, for example).
How can I best implement the code for executing the process, with a timeout? I was thinking of somehow implementing a schedule that checks if the process is still running and cancelling/destroying it if it is and the timeout has been reached.
My current (Java 8) code looks like this:
/** USE WMIC on Windows */
private static String getSystemProductUUID() {
String uuid = null;
String line;
List<String> cmd = new ArrayList<String>() {{
add("WMIC.exe"); add("csproduct"); add("get"); add("UUID");
}};
BufferedReader br = null;
Process p = null;
SimpleLogger.debug("Attempting to retrieve Windows System UUID through WMIC ...");
try {
ProcessBuilder pb = new ProcessBuilder().directory(getExecDir());
p = pb.command(cmd).start();
if (!p.waitFor(TIMEOUT, SECONDS)) { // No timeout in Java 6
throw new IOException("Timeout reached while waiting for UUID from WMIC!");
}
br = new BufferedReader(new InputStreamReader(p.getInputStream()));
while ((line = br.readLine()) != null) {
if (null != line) {
line = line.replace("\t", "").replace(" ", "");
if (!line.isEmpty() && !line.equalsIgnoreCase("UUID")) {
uuid = line.replace("-", "");
}
}
}
} catch (IOException | InterruptedException ex) {
uuid = null;
SimpleLogger.error(
"Failed to retrieve machine UUID from WMIC!" + SimpleLogger.getPrependedStackTrace(ex)
);
// ex.printStackTrace(System.err);
} finally {
if (null != br) {
try {
br.close();
} catch (IOException ex) {
SimpleLogger.warn(
"Failed to close buffered reader while retrieving machine UUID!"
);
}
if (null != p) {
p.destroy();
}
}
}
return uuid;
}
You can use the following code which only uses features available under Java 6:
public static boolean waitFor(Process p, long t, TimeUnit u) {
ScheduledExecutorService ses = Executors.newSingleThreadScheduledExecutor();
final AtomicReference<Thread> me = new AtomicReference<Thread>(Thread.currentThread());
ScheduledFuture<?> f = ses.schedule(new Runnable() {
#Override public void run() {
Thread t = me.getAndSet(null);
if(t != null) {
t.interrupt();
me.set(t);
}
}
}, t, u);
try {
p.waitFor();
return true;
}
catch(InterruptedException ex) {
return false;
}
finally {
f.cancel(true);
ses.shutdown();
// ensure that the caller doesn't get a spurious interrupt in case of bad timing
while(!me.compareAndSet(Thread.currentThread(), null)) Thread.yield();
Thread.interrupted();
}
}
Note that unlike other solutions you can find somewhere, this will perform the Process.waitFor() call within the caller’s thread, which is what you would expect when looking at the application with a monitoring tool. It also helps the performance for short running sub-processes, as the caller thread will not do much more than the Process.waitFor(), i.e. does not need to wait for the completion of background threads. Instead, what will happen in the background thead, is the interruption of the initiating thread if the timeout elapsed.
I want to perform functional test of JPOS based ISO message and response(TPS) using Jmeter.Can anyone guide me on this?Do I need to install any plugin in Jmeter.As I have tried to install the Jmeter in my system.But Can't see any Jpos plugin in listeners to start with.Dev is Using Socket connection in Jpos.
I don't think you will be able to find any plugin, however you should be able to use jPOS client library from the JSR223 Sampler
Download the jPOS and build it using Gradle
Copy jpos/build/libs/jpos-x.x.x.jar to JMeter Classpath (along with dependencies, if you don't have them yet)
Restart JMeter to pick the library up
Add JSR223 Sampler to your Test Plan and put the code, implementing your test scenario into "Script" area, an example would be something like:
import org.jpos.iso.ISOMsg
import org.jpos.iso.channel.ASCIIChannel
import org.jpos.iso.packager.ISO87APackager
def host = 'your_host'
def port = 1234
def channel = new ASCIIChannel(host, port, new ISO87APackager())
channel.connect()
def message = new ISOMsg()
message.setMTI("0800")
message.set(3, "000000")
message.set(41, "00000001")
message.set(70, "301")
channel.send(message)
ISOMsg response = channel.receive()
channel.disconnect()
More information:
jPOS Programmer’s Guide
Apache Groovy - Why and How You Should Use It
We can send a XML or rawMessage to jpos server from the Jmeter TCPSampler
You can also add TPS listeners
enter image description here
And JPosTCPClient can be implemented
public class JPosTCPClient extends TCPClientImpl {
private static final Logger log = LoggingManager.getLoggerForClass();
private String containsString = "</isomsg>";
private boolean filterEnabled = true;
public JPosTCPClient() {
filterEnabled = Boolean.parseBoolean(JMeterUtils.getPropDefault("jpos.tcp.use", "true"));
containsString = JMeterUtils.getPropDefault("jpos.tcp.contains", "</isomsg>");
}
/**
* Reads data until the defined EOL byte is reached.
* If there is no EOL byte defined, then reads until
* the end of the stream is reached.
*/
#Override
public String read(InputStream is) {
byte[] buffer = new byte[4096];
ByteArrayOutputStream w = new ByteArrayOutputStream();
int x = 0;
boolean contains = false;
try {
while ((x = is.read(buffer)) > -1) {
w.write(buffer, 0, x);
if(filterEnabled){
String response = new String(buffer);
if(response.contains(containsString)){
contains = true;
break;
} else {
System.out.println("Contents: " + response);
}
}
}
if(filterEnabled && !contains){
System.out.println("Skipped containsString checking, x length:" + x);
}
} catch (SocketTimeoutException e) {
// drop out to handle buffer
System.out.println(e.getMessage());
} catch (InterruptedIOException e) {
// drop out to handle buffer
System.out.println(e.getMessage());
} catch (IOException e) {
System.out.println(e.getMessage());
log.warn("Read error:" + e);
return "";
}
// do we need to close byte array (or flush it?)
log.debug("Read: " + w.size() + "\n" + w.toString());
return w.toString();
}
}
I have a Jersey server running locally, it exposes a SSE resource just like the examples here: https://jersey.github.io/documentation/latest/sse.html. I have a local webpack Angular app, that binds to the exposed GET endpoint and listens for data.
On the GET, I start up a thread to send notifications at regular intervals over 6-8 seconds. I don't see anything on the client UNTIL the EventOutput object is closed.
What am I doing wrong, and how can I fix this?
The server code WORKS with just a simple curl, i.e.:
curl http://localhost:8002/api/v1/notify
But on both Chrome and Safari the following code exhibits the behavior
Client (TypeScript):
this.evSource = new EventSource('http://localhost:8002/api/v1/notify');
this.evSource.addEventListener(
'event',
(x => console.log('we have ', x))
);
this.evSource.onmessage = (data => console.log(data));
this.evSource.onopen = (data => console.log(data));
this.evSource.onerror = (data => {
console.log(data);
this.evSource.close();
});
Server (Java):
// cache callback
public void eventCallback(Iterable<CacheEntryEvent<? extends Integer, ? extends Integer>> events) {
for (CacheEntryEvent<? extends Integer, ? extends Integer> x : events) {
LOGGER.info("{} Sending the following value: " + x.getValue(), Thread.currentThread().getId());
final OutboundEvent sseEvent = new OutboundEvent.Builder().name("event")
.data(Integer.class, x.getValue()).build();
this.broadcaster.broadcast(sseEvent);
}
}
#GET
#Produces(SseFeature.SERVER_SENT_EVENTS)
#ApiOperation(value = "Setup SSE pipeline", notes = "Sets up the notification pipeline for clients to access")
#ApiResponses(value = {
#ApiResponse(code = HttpURLConnection.HTTP_UNAUTHORIZED,
message = "Missing, bad or untrusted cookie"),
#ApiResponse(code = HttpURLConnection.HTTP_OK,
message = "Events streamed successfully")
})
#Timed
#ResponseMetered
public EventOutput registerNotificationEvents(
#HeaderParam(SseFeature.LAST_EVENT_ID_HEADER) String lastEventId,
#QueryParam(SseFeature.LAST_EVENT_ID_HEADER) String lastEventIdQuery) {
if (!Strings.isNullOrEmpty(lastEventId) || !Strings.isNullOrEmpty(lastEventIdQuery)) {
LOGGER.info("Found Last-Event-ID header: {}", !Strings.isNullOrEmpty(lastEventId) ? lastEventId : lastEventIdQuery );
}
LOGGER.info("{} Received request", Thread.currentThread().getId());
this.continuation = true;
final EventOutput output = new EventOutput();
broadcaster.add(output);
Random rand = new Random();
IntStream rndStream = IntStream.generate(() -> rand.nextInt(90));
List<Integer> lottery = rndStream.limit(15).boxed().collect(Collectors.toList());
IgniteCache<Integer, Integer> cache = this.ignite.cache(topic_name);
executorService.execute(() -> {
try {
lottery.forEach(value -> {
try {
TimeUnit.MILLISECONDS.sleep(500);
LOGGER.info("{} Sending the following value to Ignite: " + value + " : " + count++, Thread.currentThread().getId());
if (!cache.isClosed()) {
cache.put(1, value);
}
} catch (InterruptedException ex) {
ex.printStackTrace();
}
});
TimeUnit.MILLISECONDS.sleep(500);
continuation = false;
TimeUnit.MILLISECONDS.sleep(500);
if (!output.isClosed()) {
// THIS is where the client sees ALL the data broadcast
// in one shot
output.close();
}
} catch (InterruptedException ex) {
ex.printStackTrace();
} catch (IOException ex) {
ex.printStackTrace();
}
});
LOGGER.info("{} Completing request", Thread.currentThread().getId());
return output;
}
}
Looks like http://github.com/dropwizard/dropwizard/issues/1673 captures the problem. GZip default won't flush even if upper levels ask for it. Solution is something like
((AbstractServerFactory)configuration.getServerFactory()).getGzipFilterFactory().setSyncFlush(true);
will enable flushing to synchronize with GZip if disabling GZip all up is not an option
I have an activity that is being swapped out when I raise an intent for another activity. onPause calls saveState() to save work so far:
private void saveState() {
...
...
if (myUri == null) {
// Inserting a new record
*** myUri = getContentResolver().insert(ContentProvider.CONTENT_URI, values);
} else {
// Update an existing record
getContentResolver().update(myUri, values, null, null);
}
}
Before calling getContentResolver(), ContentProvider.CONTENT_URI = 'content://nz.co.bkd.extraTime.contentprovider/times'.
After the call, myUri = 'times/#' where #=row ID. My question is; where is the 'content:...' prefix to the returned uri?
During the call, ContentResolver.java is called and returns CreatedRow uri
ContentResolver.java
....
....
public final Uri insert(Uri url, ContentValues values)
{
IContentProvider provider = acquireProvider(url);
if (provider == null) {
throw new IllegalArgumentException("Unknown URL " + url);
}
try {
long startTime = SystemClock.uptimeMillis();
*** Uri createdRow = provider.insert(url, values);
long durationMillis = SystemClock.uptimeMillis() - startTime;
maybeLogUpdateToEventLog(durationMillis, url, "insert", null /* where */);
return createdRow;
} catch (RemoteException e) {
// Arbitrary and not worth documenting, as Activity
// Manager will kill this process shortly anyway.
return null;
} finally {
releaseProvider(provider);
}
}
At this point, createdRow = 'times/#'.
The record does actually get saved in the Sqlite database.
Do I have to add the uri prefix in my code or should the full uri be returned?
I'm using WebClient object in to poll some data from server.
It's working good and it's updating text block fine. Till I don't use map on same Page. When I add a map, only one request get completed and data is retrieved only once.
This is the code for getting messages:
public MessagesPage()
{
InitializeComponent();
new System.Threading.Timer(messagePolling, null, 0, 5000); // every 5 seconds
}
void messagePolling(object state)
{
getMessages(Const.GET_MESSAGES_URL + uuid);
}
private void getMessages(string uri)
{
WebClient webClient = new WebClient();
webClient.DownloadStringAsync(new Uri(uri));
webClient.DownloadStringCompleted += new DownloadStringCompletedEventHandler(messagesResponseCompleted);
}
void messagesResponseCompleted(object sender, DownloadStringCompletedEventArgs e)
{
lock (this)
{
try
{
string s = e.Result;
if (s.Length > 0)
{
List<Message> messagesResult = JSONHelper.Deserialize<List<Message>>(s);
foreach (Message m in messagesResult)
{
tbMessages.Text += m.message + "\n";
}
}
else
{
tbMessages.Text += "No new messages #: " + System.DateTime.Now + "\n";
}
}
catch (System.Net.WebException we)
{
MessageBox.Show(we.Message);
}
}
}
Anyone?
The WebClient response is processed on the UI thread - so you don't need the lock that you have in your event handler.
For your particular problem - is this just occurring in the emulator? I've seen quite a few timer issues with the emulator - but never anything similar on the real phone.
As an aside, I believe in general it's better to use HttpWebRequest rather than WebClient - see the explanation here of webclient using the UI thread Silverlight Background Thread using WebClient - for your particular code I don't think this will be a problem.
If using
System.Windows.Threading.DispatcherTimer myDispatcherTimer = new System.Windows.Threading.DispatcherTimer();
myDispatcherTimer.Interval = new TimeSpan(0, 0, 0, 0, 5000);
myDispatcherTimer.Tick += new EventHandler(messagePolling);
myDispatcherTimer.Start();
instead of
new System.Threading.Timer(messagePolling, null, 0, 5000); // every 5 seconds
is working fine =)