Hi
Just wanted to check with the group if anyone is experiencing unusual behaviour related to endpoints v2 on AppEngine. We are using the AppEngine standard environment.
This morning we start getting reports from users where our Android client is crashing a lot. Upon investigation we discover that yesterday (30.03.2017) at 16:36 UTC we start getting unusual entries in the log that look like this:
/_ah/api/appendpoint/v8/users/<user id omitted>/dashboarddata
and a HTTP 302 response according to the logs (this then results in a crash in the Android client, which is a different story)
On a regular day and most of the time this call from the client looks like this.
/_ah/spi/is.app.services.dashboard.DashboardEndpointV9.getDashboardData
The strange thing here is the format of the log entry which is consistent with the format of log entries after upgrading to endpoints v2.
We see the same exact Android client making this call and once in a while it seems to get the new type of response (the endpoints v2 type log entry and a 302).
We did not deploy anything at this time yesterday and we were certainly not using endpoints v2 at that time.
After some testing we have now improved the situation by upgrading our backend to using endpoints v2 (version 2.0.5).
I say improved because now most log entries look like they are coming from endpoints v2 but once in a while we get the old style log entries and now the old style log entries return 302 (according to server logs)
The Android client no longer crashes but doesn't receive a response when this happens.
To further investigate what was happening we plugged in a http proxy to see exactly what was going on.
On a successful call there is a GET call to the server with a proper HTTP 200 OK response
On a failure there is a GET call to the server with a HTTP 404 NOT FOUND response and an "unsupportedProtocol" error.
We don't see any difference between the two calls. The server logs tell us that both are served by the same instance on the backend.
So is anyone experiencing something similar?
Should we pin this on some Google update or possibly some misconfiguration on our end?