Server Unreachable Errors with NSURLSession - ios

Question

I have been using NSURLSession on my app since iOS 7 first came out. Recently some users started complaining that some features weren't working, after having a deep look at their logs I noticed that several network requests failed with error:
{NSLocalizedDescription=Server is unreachable. Check your network connection and try again., NSUnderlyingError=0x174c4c300 "The operation couldnât be completed. (NSURLErrorDomain error -1001.)"}
As I have no idea what the problem might be, I added some code to resolve the hostname when the requests failed:
NSString *urlString = [url host];
const char *hostUrl = [urlString UTF8String];
struct hostent *remote_entity = gethostbyname(host);
if (!remote_entity) {
DDLogWarn(#"DNS Lookup failed! Cant resolve Hostname: %# for request: %#", [url host], [url description]);
return nil;
}
else {
// Get address info from host entry
struct in_addr *remote_in_addr = (struct in_addr *) remote_entity->h_addr_list[0];
char *sremote_in_addr = inet_ntoa(*remote_in_addr);
NSString* hostIP = [NSString stringWithUTF8String:sremote_in_addr];
DDLogWarn(#"Hostname: %# IP: %# (original request: %#)", [url host], hostIP, [url description]);
return hostIP;
}
After deploying this change, I can see in my logs that even when the request fails ( with NSURLErrorDomain error -1001) I can still resolve the hostname.
I don't understand how this could happen, I can always resolve the hostname but the requests still fail somehow. Please note that I can't reproduce this and this bug only manifests from time to time with no predictable pattern.
Any ideas of how to proceed to fix this for good ? Should I implement a packet tracking mechanism in my apps (similar to traceroute on unix)? Could this be a misconfiguration on my servers since requests to other hosts (like Amazon S3) never fail? I'm using Heroku by the way.
Thanks in advance,
Ze

Answer 1

I recommend looking at Apple's Networking Overview Guide, specifically the Handling Network Problems Gracefully section.
When working with networking code, you have to account for the fact that sometimes things go wrong. This is especially true on mobile devices where cellular connections can come and go. In the above mentioned section on handling network problems, Apple makes recommendations for automatically retrying requests and detecting network changes using the SCNetworkReachability API. I recommend using that strategy in your networking code to give your users the best experience.
In your specific case, it appears the -1001 error is indicating a network timeout. The host lookup is succeeding, but for whatever reason, the network request did not make it to the server. In that case, a retry would make sense.