Closed xkam1x closed 1 year ago
Interesting. We seem to be dropping the connection for some reason. Given the "client dropped connection" status, I wonder if this is https://github.com/grpc/grpc-swift/pull/1626
Are you able to reproduce this easily? If so could you try building with main
?
After compiling the code against the main branch, I ran a stress test by making just under 50k RPCs and transferring just over 10GB of data. During the test, no errors were logged. This indicates that the issue, which was causing the error to be logged once every minute, has been successfully fixed.
I'm now looking forward to the next release that will incorporate this change. With the issue resolved, the upcoming release should provide a more stable and reliable experience. Thank you to the development team for their efforts in identifying and resolving the issue.
Awesome, thanks for confirming so quickly! We'll try to get the fix out soon :)
Sorry this took a little longer than hoped to get released, but it's now available in 1.19.0.
Describe the bug
Unexpectedly running into 'unavailable (14) Transport became inactive' while making RPCs. Once it happens, subsequent RPCs can also fail but system recovers in under 1 minute.
I am using the following:
My setup is APP -> NGINX -> GRPC C++ Server.
To reproduce
I am using the following channel config:
Additional information
When 'Transport became inactive' occurs NGINX logs report HTTP/2 status of 499 (Client dropped connection).
Logs just before RPC failure: