Very long request is not killed ?

[Patrice B]

Hello,

Recently, I had a request that took 40 mn to complete.  It was an ordinary http request, just with parameters that made it very long to process.  This hogged the instance and caused high latency for the whole time.

Obviously, we need to work on checking the request's parameters so that we can identify very early that this request will be too long and reject it.

However, this raises 2 questions:

1) I thought maximum response time for an http request was 1 mn, so I would have expected the request to be aborted somehow.   Maybe it was aborted for the client, but the processing did continue for 40 mn, and the request appears in the log not as an error, but as successfully processed in 40 minutes.

2) In response, the load balancing did spawn more instances, as would be expected, but strangely the latency seemed to be affected on all instances.   Can this be real, or is it just a coincidence ?

Thank you for your thoughts and expertise.   Patrice.

[Elliott (Cloud Platform Support)]

Hello Patrice,

Please note that Google Groups are reserved for general Google Cloud Platform-end product discussions and not for reporting issues, which is why I suggest moving the troubleshooting to Issue Tracker, where issues can be turned private in case we need to gather any project specific details.