I have a similar issue. After having tested my app with thousands of users it now fails (performance wise) when serving just a few users. Even if I just have 1 QPS incoming, it will boot up 10 instances and all of the loading requests are user facing (note that 1 instance would be more than enough to handle the traffic). When all those instances are busy it will boot up new ones, and it will use the warmup requests only sometimes for that.
It seems to me that the algorithm App Engine uses to determine how many instances are needed for the current traffic is somewhat flawed... On 9 Dec, 18:44, Sergio Lopes <slo...@gmail.com> wrote: > That's right, I use GAEJ. In Java, the instance concept and the > startup momento are all well defined. My problem here is that the > loading_requests rarely are /_ah/warmup requests. Most loading > requests are user requests, even with AlwaysOn enabled. It doesn't > seem right... > > On Dec 9, 3:17 pm, Stephen Johnson <onepagewo...@gmail.com> wrote: > > > > > My assumption would be and this is for Java (not a python person) that if > > you have an empty handler at the very least you would get the JVM up and > > running and some of the basic libraries that are required. Then, depending > > on what you do in the warmup request and/or your web.xml would load > > additional libraries. For example, if you use the load-on-startup with a > > integer >= 0 then the servlet would be loaded when the request handler is > > called regardless of whether or not it is used. Also, if a datastore or > > memcache request were made in the handler that those additional libraries > > would be loaded. I haven't implemented one yet but my intention is to use it > > to load as much of these additional libraries as possible. > > > On Thu, Dec 9, 2010 at 9:45 AM, Darien Caldwell > > <darien.caldw...@gmail.com>wrote: > > > > It sounds wrong. > > > > But it brings me back to a question i've been wondering to myself for > > > awhile now: what constitutes an Instance, exactly? Is it the bulk of > > > all the code uploaded, or just whatever modules were las used? > > > > For instance, I have several separate sections to my app, and > > > depending on which URL a request is received on, different code paths > > > are executed, each loading different modules. If I have a warm > > > instance, exactly what is being kept warm? All of the modules for all > > > of the URLs, or what exactly? > > > > How does the system decided what to keep warm? > > > > -- > > > 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-appeng...@googlegroups.com. > > > To unsubscribe from this group, send email to > > > google-appengine+unsubscr...@googlegroups.com<google-appengine%2Bunsubscrib > > > e...@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-appeng...@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.
