We have three small web applications hosted at discountasp.net that are callbacks invoked by a third-party API. This third-party reported intermittent timeout errors on some of the HTTP requests made to these callbacks during the last 30 days (logging information for requests made earlier is no longer available). I checked in the server logs, and for each of the failed requests, there is no corresponding entry to be found. Here are the dates and times (UTC) at which the missing requests were made: 2016-08-15 13:57:56 2016-08-15 14:24:51 2016-08-15 14:28:00 2016-08-30 15:48:28 2016-09-05 15:37:11 2016-09-06 14:01:22 2016-09-06 14:05:02 2016-09-07 18:08:48 2016-09-08 13:05:39 I saw that you have reported some networking issues both on August 15 and 30, and also on September 8, although none on September 5, 6, and 7. Were this network issues such as to cause some requests to be lost? If so, were there any network hiccups on the other dates?
Anything that would affect the network (or an individual server for that matter) would be posted in the outages and maintenance forum. A timeout error could be caused by a lot of things, including the application resetting due to an application pool time out (caused by inactivity) or connection problems outside of our network. Difficult to troubleshoot unless it's happening consistently...
Thanks. It certainly does not happen consistently. On 2016-09-08 for example, 439 similar requests were posted, only one went missing. I must admit I did not observe any noticeable gaps around the times when the missing requests were sent that could be indicate a network unavailability. On 2016-09-08, there were requests being processed by a web service sitting on the same server around the time the missing request should have been received. On the other hand, the third-party support staff says: "[the 2016-09-08 request] shows a 15-second timeout on our side, which indicates we successfully connected to your server, but we did not receive an HTTP response within the allotted time."
If it's possible that there were 20 minutes of inactivity, that would make sense, since the application has to spin up after a timeout, and that could take more than 15 seconds, depending on the size of your application. But again, the lack of a response within a specified amount of time can have a lot of different causes, so it might be worth looking at the application itself, to make sure there aren't times where it's lagging due to being busy or an excess of concurrent connections...it's a lot easier to troubleshoot if there is some consistency, or if it happens during certain times of day, etc.
And in that case (application pool shut down due to long inactivity), assuming that it takes longer than 15 seconds before the application is ready, the incoming request would be dropped if sender (third-party) closes the connection? That request would not appear in the IIS log? Yes, I agree it is not easy to troubleshoot, especially that there is a third player here. BTW, it happened again yesterday at 2016-09-13 22:55:26 UTC. This time, the log shows an IIS header inserted at that time. 2016-09-13 22:52:17 ... POST ... #Software: Microsoft Internet Information Services 7.5 #Version: 1.0 #Date: 2016-09-13 22:56:54 #Fields: ... 2016-09-13 22:56:53 ...POST ... I did not however see anything like this in the previous logs.
