Yes, this problem is very annoying. I think it is a defect even a bug in the instance scheduler. Below is mail I sent to GAE support, without replies. The problem is the scheduler often will not use an instance in idling.
***************************************** This problem is so serious. Above 30% percent of my app requests will be slow loading. I hope you can fix this problem soon. Here I provide a detailed example for this problem: ------------------------------------------------- 1. Instance Snapshots at 8:16 Instances help QPS* Latency* Requests Errors Age Memory Availability 0.050 14589.0 ms 1 0 0:00:26 80.8 MBytes Dynamic Icon Dynamic 0.000 0.0 ms 2 0 0:31:51 80.0 MBytes Dynamic Icon Dynamic 2. Instance Snapshots at 8:25 Instances help QPS* Latency* Requests Errors Age Memory Availability 0.050 11925.0 ms 1 0 0:00:27 75.7 MBytes Dynamic Icon Dynamic 0.000 0.0 ms 2 0 0:41:08 80.0 MBytes Dynamic Icon Dynamic 3. Instance Snapshots at 8:27 Instances help QPS* Latency* Requests Errors Age Memory Availability 0.000 0.0 ms 2 0 0:42:38 80.0 MBytes Dynamic Icon Dynamic -------------------------------------------------- You will find at 8:25, a new instance is created with 12 seconds loading and handle a new request. But the number handled requests for the other instance is 2, which is the same as 9 minutes ago. This proves when the new instance is being created, the other long run instance is idle and available. It is unnecessary to create the new instance. Even if the new instance should be created for other reasons, the scheduler should still let the long run idle instance to handle the just coming request, so that the visitor doesn't need to wait for 12 seconds to load a page. On Nov 18, 4:01 pm, nicois <[email protected]> wrote: > I'm developing another app at the moment, using the latest python SDK, > python 2.7 with threading enabled, and I'm seeing a few strange > things: > > The dashboard reports 9 requests, but the event log shows 61 in the > last 24 hours. > The maximum CPU time used on a request is under 10000ms. > The number of instance hours used in the same period is 1.36, out of > 28 available (ie:~5%). > > These statistics are my main concern. By my calculations, even > assuming each request took the full 10 seconds, I should have used > about 10 CPU minutes, rather than the 100 I'm seeing. > > The second thing which I am experiencing today, and which may or may > not be related to the above: if I do not make a request for 10-20 > seconds, the GAE instance is shut down, requiring a complete reload of > the instance at the next request. This means if someone loads a page, > they have to wait ~10seconds, and after digesting that page, if more > than a few seconds have passed, they will have another 10 second wait > for the next page. > > I can understand GAE spinning down instances when not required, but > this seems to be occurring very agressively, and is probably resulting > in an increase in the CPU load being put on the GAE infrastructure. > > Has anyone else experienced anything like this with their apps? I > couldn't see any issues in the GAE issue tracker, and wanted to make > sure I wasn't doing anything silly before logging a ticket there. > > thanks. -- You received this message because you are subscribed to the Google Groups "Google App Engine" group. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/google-appengine?hl=en.
