[google-appengine] Re: Latency problem still really a problem
They said it has been "stabilized." But I agree, it is still much worse. The timeouts are the real problem; it's just impossible to complete certain operations. On Mar 6, 3:05 pm, cz wrote: > Our app is still experiencing at least double the latency as compared > to last week and the app is basically unusable. I'm really concerned > about this since the GAE team has stated that the latency issue has > been resolved. > This is so disappointing if the current performance is considered to > be fine. I'm really hoping that this isn't the case. > > I'm curious as to how you are testing latency internally? Perhaps the > tests aren't reflecting very well some real world scenarios and that > some combination of datastore operations are particularly onerous. > > In any case, latency is still much much worse than before the server > maintenance. > > P.S. also getting random 502 errors which I assume is related to this > issue:http://groups.google.com/group/google-appengine/browse_thread/thread/... --~--~-~--~~~---~--~~ 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 -~--~~~~--~~--~--~---
[google-appengine] Re: Latency problem still really a problem
I've got stable performance in my application. It consistently runs 3 to 4 times slower than it used to. Apparently, application performance should improve on Monday but in the mean time performance is quite poor... http://code.google.com/status/appengine On Mar 6, 8:05 pm, cz wrote: > Our app is still experiencing at least double the latency as compared > to last week and the app is basically unusable. I'm really concerned > about this since the GAE team has stated that the latency issue has > been resolved. > This is so disappointing if the current performance is considered to > be fine. I'm really hoping that this isn't the case. > > I'm curious as to how you are testing latency internally? Perhaps the > tests aren't reflecting very well some real world scenarios and that > some combination of datastore operations are particularly onerous. > > In any case, latency is still much much worse than before the server > maintenance. > > P.S. also getting random 502 errors which I assume is related to this > issue:http://groups.google.com/group/google-appengine/browse_thread/thread/... --~--~-~--~~~---~--~~ 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 -~--~~~~--~~--~--~---
[google-appengine] Re: Latency problem still really a problem
On Mar 6, 7:14 pm, Brandon Thomson wrote: > They said it has been "stabilized." But I agree, it is still much > worse. The timeouts are the real problem; it's just impossible to > complete certain operations. I get timeouts trying to put single entities into the datastore (sometimes). These aren't big objects -- maybe a total size of 3k bytes with 3 indexes. I also get timeouts just paging through result sets -- i.e. when doing 'for (doc in doclist):' where doclist is the result of filter operation. It clearly is doing something when these timeouts happen as the CPU time used is *much* higher than normal. Alternatively, maybe there is a bug, and something goes into a hard loop that is only broken when the timeout expires... It would be nice to know whether the CPU that we are being charged in these timeout cases is *real* CPU that was actually consumed somewhere, or whether this is a notional CPU charge for what "must have been happening -- but actually wasn't". If the former, then I'd like to understand what was actually going on to consume all of these CPU cycles. Philip --~--~-~--~~~---~--~~ 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 -~--~~~~--~~--~--~---
[google-appengine] Re: Latency problem still really a problem
We have been experiencing datastore timeouts and, even worse, 502 timeouts since last week. A few CPU intensive pages and queries that were completing 100% of times until recently, now they constantly go into "deadline exceeded error" (after 30 seconds, when it would take only 1-2 seconds earlier) and return 502 errors. Furthermore, GAE is sensibly slower on all requests. I was surprised to see that on the status page, the google team says that "latency issues stabilized". That is not what I have experienced. On Mar 8, 6:07 pm, WeatherPhilip wrote: > On Mar 6, 7:14 pm, Brandon Thomson wrote: > > > They said it has been "stabilized." But I agree, it is still much > > worse. The timeouts are the real problem; it's just impossible to > > complete certain operations. > > I get timeouts trying to put single entities into the datastore > (sometimes). These aren't big objects -- maybe a total size of 3k > bytes with 3 indexes. I also get timeouts just paging through result > sets -- i.e. when doing 'for (doc in doclist):' where doclist is > the result of filter operation. > > It clearly is doing something when these timeouts happen as the CPU > time used is *much* higher than normal. Alternatively, maybe there is > a bug, and something goes into a hard loop that is only broken when > the timeout expires... It would be nice to know whether the CPU > that we are being charged in these timeout cases is *real* CPU that > was actually consumed somewhere, or whether this is a notional CPU > charge for what "must have been happening -- but actually wasn't". If > the former, then I'd like to understand what was actually going on to > consume all of these CPU cycles. > > Philip --~--~-~--~~~---~--~~ 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 -~--~~~~--~~--~--~---
[google-appengine] Re: Latency problem still really a problem
I suspect that it is "real CPU" that is unfortunately not doing any useful work. For example when you commit a transaction on a highly contentious entity, you may need to retry 50 or more times before succeeding. Each try wastes the same amount of CPU, but you only succeed in committing the transaction once. It's weird, I know. As to why so much CPU is being wasted now is anybody's guess; Google won't tell us what they changed. On Mar 8, 1:07 pm, WeatherPhilip wrote: > It clearly is doing something when these timeouts happen as the CPU > time used is *much* higher than normal. Alternatively, maybe there is > a bug, and something goes into a hard loop that is only broken when > the timeout expires... It would be nice to know whether the CPU > that we are being charged in these timeout cases is *real* CPU that > was actually consumed somewhere, or whether this is a notional CPU > charge for what "must have been happening -- but actually wasn't". If > the former, then I'd like to understand what was actually going on to > consume all of these CPU cycles. --~--~-~--~~~---~--~~ 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 -~--~~~~--~~--~--~---
[google-appengine] Re: Latency problem still really a problem
hi, i'm wondering how bad your performance problems are - i'm new to google app engine, just deployed my new app last weekend for the first time.. and it was _always_ slow .. but i have no idea why .. for example a page with maybe one datastore query needs 3274ms-cpu according to the logs in the administration console .. i have added the profiler which outputs: '11288 function calls (11000 primitive calls) in 0.401 CPU seconds' for the same request - so why is app engine requiring 3.2 secs, while my app "only" takes 0.4secs (for an almost static page this is still bad) can anyone tell me if this is normal speed? what can i expect? thanks, herbert --~--~-~--~~~---~--~~ 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 -~--~~~~--~~--~--~---