When we migrate our app to auction from rackspace, we have time to read about 50% of HIT requests saw.
We tried to keep the client inside the blue with the same result and outside, in this case the customer is also a server BTW, so neither the issues of any geographic / browser. We also tried to increase the size of the box to ensure that the blue was not throttling. But while using the D box for a single request, the result was the same.
Once we came out of the auction, they started working properly.
Each question was done directly on an example using a public IP, so there is no load balancer problem either.
Approximately 50% of the time the timeout ran in this problem was set for 15 minutes.
50% of HTTP requests end when the area was US 2
Doing is not common practice is the reason why you need to analyze that by validating the requests due to the deadlines that are killing your VM, I advise you to run a packet capture on your server and Analysis of reaction times, as well as the number of high sensibilities I also see; It is also better that you can take a network trace simultaneously on your client machines so that you can analyze the TCP sequence number and get a comparison of the packet obtained. If you are seeing latency in packet capture or high number of retranslation, then it needs a detailed analysis. I strongly suggest that you open a support incident so that Microsoft support can assist you in examining your problem. Could.
No comments:
Post a Comment