At the outset, to save the server, MongoDB and Redis are placed on a server. Every peak of the website special card, also often appear 502. For a long time to find the reason, found that hard disk writing data is very large, IOPS is also very high, troubleshooting a number of reasons have not been found. Then carefully study the monitoring, found that the operation of the hard disk is very regular, every few minutes there is a frequent write hard disk, Lenovo to the Redis synchronous data to the hard disk interval is a few minutes, so began to suspect that Redis caused. So added a server, put Redis alone there, found that the site instantly, 502 problems no longer appear, it is a painful experience ah. As for the scenario of placing MongoDB and Redis on the same server with different hard drives, I have not tested it, and I think it should also have an impact on MongoDB, because Redis synchronizes data to the hard disk when it is more CPU-intensive. Redis server, is a quad-core CPU, when synchronizing data to the hard disk, CPU utilization of up to 60%, perhaps also with hard disk write performance is related to it, Redis very eat memory, very occupy hard disk IO resources.