|
|
After sometime (timeout): "client disconnected during delivery of response for <WORequest[0x0x7f73209d30d8]: method=POST......the socket was shutdown". "frozen" connections each time I refreshing mail Inbox.
/nginx/
proxy_connect_timeout 60;
proxy_send_timeout 360;
proxy_read_timeout 360;
/ SOGo /
SOGoMaximumPingInterval = 354;
SOGoMaximumSyncInterval = 354;
SOGoInternalSyncInterval = 3; |
|
|
|
Also (maybe not related and maybe even not a bug) after receiving mail there are two "Change detected during Sync, we push the content." records in logs and two POSTs. Should it fire two times on single incoming message or not? |
|
|
|
By the way, after receiving mail it stops SOGoActiveSyncDispatcher, so no new mail delivered to iPhone. After some time it starts dispatch it again. Too many points, let's diagnose step-by-step starting from my first problem described in "Description". If other problems will exists after fixing it then I will open new issues there. Thanks. |
|
tfu
2016-03-28 16:08
reporter
|
3603.diff (2,784 bytes)
diff --git a/ActiveSync/SOGoActiveSyncDispatcher.m b/ActiveSync/SOGoActiveSyncDispatcher.m
index 17599ae..091efc6 100644
--- a/ActiveSync/SOGoActiveSyncDispatcher.m
+++ b/ActiveSync/SOGoActiveSyncDispatcher.m
@@ -2164,7 +2164,7 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- (void) processPing: (id <DOMElement>) theDocumentElement
inResponse: (WOResponse *) theResponse
{
- NSString *collectionId, *realCollectionId, *syncKey;
+ NSString *collectionId, *realCollectionId, *syncKey, *key, *processIdentifier;
NSMutableArray *foldersWithChanges, *allFoldersID;
SOGoMicrosoftActiveSyncFolderType folderType;
NSMutableDictionary *folderMetadata;
@@ -2177,8 +2177,13 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
NSData *d;
NSAutoreleasePool *pool;
- int i, j, heartbeatInterval, defaultInterval, internalInterval, status;
+ int i, j, heartbeatInterval, defaultInterval, internalInterval, status, total_sleep;
+ // Let other ping requests know that a new request has arrived.
+ processIdentifier = [NSString stringWithFormat: @"%d", [[NSProcessInfo processInfo] processIdentifier]];
+ key = [NSString stringWithFormat: @"%@+%@", [[context activeUser] login], [context objectForKey: @"DeviceType"]];
+ [[SOGoCache sharedCache] setValue: processIdentifier forKey: key];
+
defaults = [SOGoSystemDefaults sharedSystemDefaults];
defaultInterval = [defaults maximumPingInterval];
internalInterval = [defaults internalSyncInterval];
@@ -2272,8 +2277,29 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
}
else
{
- [self logWithFormat: @"Sleeping %d seconds while detecting changes in Ping...", internalInterval];
- sleep(internalInterval);
+ total_sleep = 0;
+
+ while (total_sleep < internalInterval)
+ {
+ // We check if we must break the current ping request since an other ping request
+ // has just arrived.
+ if ([[SOGoCache sharedCache] valueForKey: key] && ![[[SOGoCache sharedCache] valueForKey: key] isEqualToString: processIdentifier])
+ {
+ if (debugOn)
+ [self logWithFormat: @"EAS - Ping request canceled (%@)", [[SOGoCache sharedCache] valueForKey: key]];
+
+ // Make sure we end the heardbeat-loop.
+ heartbeatInterval = internalInterval = 1;
+
+ break;
+ }
+ else
+ {
+ [self logWithFormat: @"Sleeping %d seconds while detecting changes in Ping...", internalInterval-total_sleep];
+ sleep(5);
+ total_sleep += 5;
+ }
+ }
}
}
|
|
|
|
Attached patch should avoid duplicate ping requests.
-> I don't have an iPhone to test. |
|
|
|
Why not do the same thing we do for Sync requests? Using only the SOGoCache isn't cluster safe.
Using SOGoCacheGCSObject is. |
|
tfu
2016-03-28 19:31
reporter
|
3603-1.diff (3,074 bytes)
diff --git a/ActiveSync/SOGoActiveSyncDispatcher.m b/ActiveSync/SOGoActiveSyncDispatcher.m
index 17599ae..3c8a66e 100644
--- a/ActiveSync/SOGoActiveSyncDispatcher.m
+++ b/ActiveSync/SOGoActiveSyncDispatcher.m
@@ -2164,11 +2164,12 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- (void) processPing: (id <DOMElement>) theDocumentElement
inResponse: (WOResponse *) theResponse
{
- NSString *collectionId, *realCollectionId, *syncKey;
+ NSString *collectionId, *realCollectionId, *syncKey, *processIdentifier, *pingRequestInCache;
NSMutableArray *foldersWithChanges, *allFoldersID;
SOGoMicrosoftActiveSyncFolderType folderType;
NSMutableDictionary *folderMetadata;
SOGoSystemDefaults *defaults;
+ SOGoCacheGCSObject *o;
id <DOMElement> aCollection;
NSArray *allCollections;
@@ -2177,8 +2178,17 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
NSData *d;
NSAutoreleasePool *pool;
- int i, j, heartbeatInterval, defaultInterval, internalInterval, status;
+ int i, j, heartbeatInterval, defaultInterval, internalInterval, status, total_sleep;
+ // Let other ping requests know that a new request has arrived.
+ processIdentifier = [NSString stringWithFormat: @"%d", [[NSProcessInfo processInfo] processIdentifier]];
+ o = [SOGoCacheGCSObject objectWithName: [context objectForKey: @"DeviceId"] inContainer: nil useCache: NO];
+ [o setObjectType: ActiveSyncGlobalCacheObject];
+ [o setTableUrl: [self folderTableURL]];
+ [o reloadIfNeeded];
+ [[o properties] setObject: processIdentifier forKey: @"PingRequest"];
+ [o save];
+
defaults = [SOGoSystemDefaults sharedSystemDefaults];
defaultInterval = [defaults maximumPingInterval];
internalInterval = [defaults internalSyncInterval];
@@ -2272,8 +2282,30 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
}
else
{
- [self logWithFormat: @"Sleeping %d seconds while detecting changes in Ping...", internalInterval];
- sleep(internalInterval);
+ total_sleep = 0;
+
+ while (total_sleep < internalInterval)
+ {
+ // We check if we must break the current ping request since an other ping request
+ // has just arrived.
+ pingRequestInCache = [[self globalMetadataForDevice] objectForKey: @"PingRequest"];
+ if (pingRequestInCache && ![pingRequestInCache isEqualToString: processIdentifier])
+ {
+ if (debugOn)
+ [self logWithFormat: @"EAS - Ping request canceled (%@)", pingRequestInCache];
+
+ // Make sure we end the heardbeat-loop.
+ internalInterval = heartbeatInterval;
+
+ break;
+ }
+ else
+ {
+ [self logWithFormat: @"Sleeping %d seconds while detecting changes in Ping...", internalInterval-total_sleep];
+ sleep(5);
+ total_sleep += 5;
+ }
+ }
}
}
|
|
|
|
Attached a new version of the patch. |
|
|
|
How about cleaning up this data in case we got stale data with sogo-tool?
You could also replace in your patch:
processIdentifier = [NSString stringWithFormat: @"%d", [[NSProcessInfo processInfo] processIdentifier]];
o = [SOGoCacheGCSObject objectWithName: [context objectForKey: @"DeviceId"] inContainer: nil useCache: NO];
[o setObjectType: ActiveSyncGlobalCacheObject];
[o setTableUrl: [self folderTableURL]];
[o reloadIfNeeded];
with:
o = [self globalMetadataForDevice]; |
|
|
|
I'd love to test! Can I wait for nightly build or need to compile it myself? |
|
|
|
I don't think that there is an additional cleanup needed. "globalMetadata" will be cleaned with resetdevice.
With the change you suggested o would be a dictionary, how do you suggest to do the save ([o save])? |
|
|
|
[o save] :-) |
|
|
|
don't think so.
[self globalMetadataForDevice] doesn't return a SOGoCacheGCSObject. |
|
|
|
Ah, true - forget it then. |
|
|
|
Also fixed for 2.3.10. |
|
