iOS: NSURLConnection callbacks heavily delayed or not fired at all - ios

Question

I got a VoIP iOS App using a long polling mechanism to maintain its service connections and receive events (calls). This means, an NSURLConnection is pending for several minutes but will return immediately after an event occured. Due to the VoIP flag, it is possible to setup a keep alive handler and receive updates even while the app is in background mode.
However, this works most of the time but not reliably. Sometimes, NSURLConnection callbacks are heavily delayed or not fired at all, even after the request timed out (timeoutInterval of the NSURLRequest reached).
An example from the log to clarify:
The app runs in background mode (is launched by the system at boot time)
NSURLConnection #1 (long poll) is initiated and returns after 1 minute with some new data
NSURLConnection #2 (long poll) is initiated and returns after 15 minuted (server side maximum) without any new data
(...)
NSURLConnection #99 (long poll) is initiated but does NOT return - not even after timeoutInterval expired (16 minutes)
From time to time, the keep alive handler is called, bot nothing happens. The backgroundTimeRemaining property got an unrealistic high value (179769313486231570814527423731704356798070567525844996598917476803157260780028538760589558632766878171540458953514382464234321326889464182768467546703537516986049910576551282076245490090389328944075868508455133942304583236903222948165808559332123348274797826204144723168738177180919299881250404026184124858368.0 instead of a maximum of 180.0).
After 1 hour, the user opens the app. The app is able to execute various NSURLRequests and receivce responses
After some secounds, the user closes the app
After 10 more minutes, the NSURLConnection #99 callback didFailWithError got fired with a timeout error (-1001). The exceution time of this request was more than one hour even tough the timeoutInterval was limited to 16 minutes and several other requests where initated later but completed earlier.
From my point of view, this seems like a very strange behaviour of iOS. Why should iOS give background excution time to the app and call the keep alive handler, but not properly fire the NSURLConnection callbacks in time?
Keep alive handler:
- (BOOL)application:(UIApplication *)application didFinishLaunchingWithOptions:(NSDictionary *)launchOptions {
[[UIApplication sharedApplication] setKeepAliveTimeout:600 handler:^{
NSLog(#"########### Started Keep-Alive Handler ###########");
[self startBackgroundHandler:YES timeout:30];
NSLog(#"########### Completed Keep-Alive Handler ###########");
}];
[self startBackgroundHandler:NO timeout:60];
}
-(void)startBackgroundHandler:(BOOL)force timeout:(int)timeout {
UIApplicationState currentAppState = [[UIApplication sharedApplication] applicationState];
BOOL appIsBackground = currentAppState == UIApplicationStateBackground;
if(appIsBackground || force) {
int localThreadId = ++_currentBackgroundThreadId;
__block UIBackgroundTaskIdentifier bgTask;
bgTask = [[UIApplication sharedApplication] beginBackgroundTaskWithExpirationHandler:^{
// Clean up any unfinished task business by marking where you
// stopped or ending the task outright.
NSLog(#"Cleaning up [Background Thread %d] ...", localThreadId);
[[UIApplication sharedApplication] endBackgroundTask:bgTask];
bgTask = UIBackgroundTaskInvalid;
}];
NSLog(#"startBackgroundHandler with [Background Thread %d] appIsBackground=%d force=%d", localThreadId, appIsBackground, force);
// Start the long-running task and return immediately.
dispatch_async(dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0), ^{
if(_currentBackgroundThreadId == localThreadId) {
NSLog(#"[Background Thread %d] Background time left: %0.1f", localThreadId, [UIApplication
sharedApplication].backgroundTimeRemaining);
sleep(timeout);
}
NSLog(#"[Background Thread %d] Will exit...", localThreadId);
[[UIApplication sharedApplication] endBackgroundTask:bgTask];
bgTask = UIBackgroundTaskInvalid;
});
} else {
NSLog(#"Ignored startBackgroundHandler - appIsBackground=%d force=%d", appIsBackground, force);
}
}
All NSURLConnections have a runloop - they are started as follows:
NSURLConnection* connection = [[NSURLConnection alloc] initWithRequest:mrequest delegate:self startImmediately:NO];
if(connection) {
[connection scheduleInRunLoop:[NSRunLoop mainRunLoop] forMode:NSDefaultRunLoopMode];
[connection start];
} else {
// error handling...
}
PS:
In an earlier version of the app, is used the data fetch background mode instead of voip, and never ran into that kind of problem.

Answer 1

Seems like iOS is not willing to give you CPU time on the main thread while you are in the background - even if you got the VoIP flag. Therfore, you should schedule your request in a separate thread, and use CFRunLoopRun() to make it run in the background. Additionally, you have to trigger execution the appropriate mode, using runMode:beforeDate: on the runloop.
BUT the real problem is that iOS will stop giving CPU time if the requests got timeouts too often - you really need to receive anything in order to get CPU time. Therefore, it is important to have the WebServer replying in time.