The only reason I was not completely dead in the water during this time is many of the pages can be served directly from memcache without loading the whole stack, so typically the front page would respond in 20ms but woe betide hitting any page that wasn't in the stack.
My testing is specifically targetted at loading the stack when no page is in cache and I was typically getting killed with DeadlineExceeded before I could finish importing stuff (yep I working feverishly on reducing this startup time by deferring things to later and only if needed and reducing the complexity of the configuration/registration process of the components,) T On Nov 26, 6:57 am, Tim Hoffman <[email protected]> wrote: > The instance with problems seems to have settled down. > > The period of high numbers of DeadLineExceeded ran from just around > the time statistic got reset through too around 5:40am (google time) > I define google time as the time stamp in the logs. > > Which is a very long period of extreme startup times. (I working hard > on reducing my startup times as much as possible) and I believe I can > reduce > them by a fair amount in a normally running system > > However to have two instances running the same code release and > essentially performing the exact same task (that I am monitoring) > and one consistently doing a full stack load in <= 10sec vs consistent > DeadlineExceeded error is a real worry because it means to > me I can get an instance caught in a performance backwater that > googles main monitoring tools (like the system status) are probably > not going to > see at all. > > T > > On Nov 25, 9:53 pm, Tim Hoffman <[email protected]> wrote: > > > I have also noted that the dashboard for polytechnic-wa is being > > updated > > Specifically the Resource Summary, Current load and Errors > > sections. > > > The other instance is right up to date. > > > There definately seems to be something wrong with my polytechnic-wa > > instance. > > (I have reployed a couple of times to see if that would make any > > difference) > > > T > > > On Nov 25, 8:33 pm, Tim Hoffman <[email protected]> wrote: > > > > Hi > > > > I have two app instances, one is being used for training and testing > > > and the other is in production (or it should be if it wasn't > > > DeadlineExceeding all the time). > > > > I am monitoring both and performing a simple diagnostics get which > > > does load the entire stack but does no real work. > > > > The training one (psc-dev1) is consistently posting response times for > > > a complete load of under 2sec and sometimes as high as 10sec > > > > The production one polytechnic-wa during the same period is almost > > > consistantly getting DeadlineExceeded errors very early into the > > > startup (just after some inital imports and just starting main) > > > > Its a big concern to me that two instances running the same code can > > > show such wildly different performance. > > > > Can anyone from google have a look at this, (is one of my instances > > > stuck on some flaky hardware for instance) > > > > Thanks > > > > Tim Hoffman -- 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.
