While scanning a large table in GCS it is possible to trigger a deadlock.
This seems to be unique to GCS & HTTP1, possibly because this GCS multiplexes over many TCP connections (vs AWS which multiplexes over a single TCP connection). Once in this deadlock no futures are getting polled.
While scanning a large table in GCS it is possible to trigger a deadlock.
This seems to be unique to GCS & HTTP1, possibly because this GCS multiplexes over many TCP connections (vs AWS which multiplexes over a single TCP connection). Once in this deadlock no futures are getting polled.