Currently, the RSCoordinator background thread parses N times the same command, with N as shard number, making it waste a significant portion of CPU cycles on it. On the bellow example ( with 25 shards ) it took 8.5% of the total background thread CPU cycles.
Currently, the RSCoordinator background thread parses N times the same command, with N as shard number, making it waste a significant portion of CPU cycles on it. On the bellow example ( with 25 shards ) it took 8.5% of the total background thread CPU cycles.