Hello, we faced the following problem: when one shard with mongod becomes unavailable ( connection is refused, e.g. a box is powered off; when mongos can't get some response from mongod within the adequate time), all mongoses from a shard cluster begin slowing down. It makes all mongos wait for some response from the dead shard. As result, all mongoses start timouting. This patch introduces a new parameter for mongos which allows to set a timeout on the connection between mongoses and shards.
Hello, we faced the following problem: when one shard with mongod becomes unavailable ( connection is refused, e.g. a box is powered off; when mongos can't get some response from mongod within the adequate time), all mongoses from a shard cluster begin slowing down. It makes all mongos wait for some response from the dead shard. As result, all mongoses start timouting. This patch introduces a new parameter for mongos which allows to set a timeout on the connection between mongoses and shards.