I believe most who suffer from this problem understand the intended operation of the scheduler as stated in the docs.
Some just imply that the design is sub-optimal. A much louder complaint is the lack of consistent behavior, whether it is understood or not. David On Thursday, March 7, 2013 1:13:46 AM UTC-7, Kristopher Giesing wrote: > Be careful how you analyze the situation with your instances. When I was > playing with resident instances, I assumed that they represented a "floor" > you wouldn't drop below in times of low traffic, but that's not quite what > they do. Rather, they represent a pool of *deliberately idle* resources that > can absorb spikes. > > > When working according to Google's intent, resident instances will absorb > incoming traffic, but as soon as they become busy (with even one request) > dynamic instances will spin up to offset the loss of idle instances. Since > this happens very close together, it's easy to assume that the dynamic > instance was spun up to handle a given request and that the resident instance > got nothing. But in my particular case, what happened was that the resident > instance got the first request, concurrently with the warmup for the dynamic > request, and then all subsequent requests went to the dynamic instance. > > > Once I understood what resident instances do, I realized they do not - by > design - solve the problem I wanted to solve, and I stopped using them > altogether. I never actually produced evidence to demonstrate that traffic > was going to a cold instance when a resident instance was available. I > believe others when they say they have such evidence, but I also strongly > suspect that Google lumps all such claims into the same "not understanding > what resident instances are supposed to do" bucket and ignores them. > > > - Kris > > On Wednesday, March 6, 2013 5:45:30 PM UTC-8, Cesium wrote:Tapir, > > > > This is an ongoing issue we have suffered with. > > > > Get used to it. There will be no (useful) response from G. > > > > David > > > > On Tuesday, March 5, 2013 6:52:28 PM UTC-7, Tapir wrote: > > > There is a resident instance obviously there. But the GAE scheduler often > > ignores it totally and always creates a new instance then the let the new > > instacne to handle the new request. > > > What is the functionality of the resident instance? If the scheduler always > > creates a new instance, the why put a resident instance there? > > > > > > It really make the user experience very bad. -- You received this message because you are subscribed to the Google Groups "Google App Engine" group. To unsubscribe from this group and stop receiving emails from it, send an email to google-appengine+unsubscr...@googlegroups.com. To post to this group, send email to google-appengine@googlegroups.com. Visit this group at http://groups.google.com/group/google-appengine?hl=en. For more options, visit https://groups.google.com/groups/opt_out.
