Request was aborted after waiting too long to attempt to service your request

[o...@haitov.com]

Hi,

I'm experiencing a very strange issue with my app, App engine suddenly can't deal with incoming requests and shows this message in the logs without any error code and at information level ! 

Sometimes all the incoming requests during a few minutes can return this message and then everything goes back to normal !

This is really bad as it is a front end that serves a mobile app and that means that during this time the app is fully non functional :-(((

@Google 

Why isn't this message at error or warning level ? 

Attached is a screenshot of the log message

[Jason Collins]

Requests that hit your application are put into a queue waiting for an instance to become available. If those requests become 10-seconds old, they are cancelled (error to client).

The App Engine Scheduler may or may not attempt to spin up a new instance under this pressure (rules are black box and change from time to time). If it is able to spin up an instance, it will send the request to the "cold" (no warm up request) instance, though there is much debate (e.g., https://code.google.com/p/googleappengine/issues/detail?id=7865) about whether or not this should occur, especially for Java apps.

The only way to avoid these errors are to allocate Min Idle Instances, which keeps resident instances around to help in this specific case. These resident instances themselves can be somewhat confusing because they are really only used to serve when a dynamic instance cannot be found - i.e., basically under this pressure situation you've outlined. You'll find, in a more smoothly loaded case, that the resident instances can be very under-utilized as they are basically idle waiting for spikes. This too is the subject of much debate.

j