Hi, This has generally beed a recurring issue with both master-slave apps and django. It is very possibly hanging up on your imports. You might add some logging to verify that, but it is commonly the case.
Migrating to high-replication could help with this. Dropping django can be a further boost. Robert On Fri, Jan 20, 2012 at 04:45, UseShots <goo...@useshots.com> wrote: > Hi Robert, > > > It's a three year old app that I didn't migrate to a high replication > datastore (datastore use is really minimum though, I just periodically save > cache values for persistence - less than 1% of requests use datastore ) > > >>Also, you note that pages typically take 1-4 seconds to load? That seems >> very slow. > > 1-4 seconds is for new instance loads. Typical request to warm instances > takes about 0.2 - 0.5 seconds. Some requests take several seconds but only > because they involve urlfetch to third-party sites. And in this thread I'm > not talking about those urlfetch requests (they have less that 10s timeout > anyway). > > The problem is with plain pages that only load django templates. I'm quite > happy with their normal performance (see above), but sometimes I observe > weirds periods when new instances take about a minute to load. And > yesterday, the situation became even worse when GAE tried to open new > instances that would just freeze for about 10 minutes (still spending my > instance-hour quota). With those frozen instances, GAE tried to launch more > and more new instances that would just freeze too. Luckily, that was a > pretty slow part of the day and there was not many requests to my app so > during the two peaks I had 15-20 active instances (where 1 instance is > enough). As a result, those to peaks have almost exhausted my app's free > instance-hour quota. It happened for the first time, and there were no > spikes in requests (I would say the request/second ratio was below the > average). > > What's really interesting, is that some request would return the 200 > response code after that long (~10minutes) execution times. How is this > possible? I guess it was some error in computing execution times during that > period as they didn't correlate to the times I saw for my own debug > messages: I have debug messages in the beginning of requests and in the very > end - the difference between the start and end times is approximately the > same as the reported execution time. But during those periods of increased > latency, there is no correlation between my calculated time and the reported > execution time - sometimes they differ by orders of magnitude. Moreover, > sometimes I noticed that the reported execution time is longer that the > observed execution time. And this is really strange as that time counts > toward the instance-hour quota (as far as I can tell) > > I've attached some screenshots to show the "request/second", "instances" and > "milliseconds/request" charts. > > -- > You received this message because you are subscribed to the Google Groups > "Google App Engine" group. > To view this discussion on the web visit > https://groups.google.com/d/msg/google-appengine/-/1u4WJKzDFHMJ. > > To post to this group, send email to google-appengine@googlegroups.com. > To unsubscribe from this group, send email to > google-appengine+unsubscr...@googlegroups.com. > For more options, visit this group at > http://groups.google.com/group/google-appengine?hl=en. -- You received this message because you are subscribed to the Google Groups "Google App Engine" group. To post to this group, send email to google-appengine@googlegroups.com. To unsubscribe from this group, send email to google-appengine+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/google-appengine?hl=en.