In this chapter, you learned the basic reasons for MongoDB failures and the situations in which they might happen. We learned that the 32-bit version of MongoDB has some issues in the production environment, and it's necessary to use a 64-bit version for production only to prevent issues in high load.

After this, we learned about the ulimit function in Unix family operating systems, which enable us to manage limitations for processes or change the limitation easily via the command-line environment. Most of the time, MongoDB can reach this limitation and crashes due to not having enough resources.

We then went on to discover that while replica sets are available for failover and the election process, MongoDB is not available for write operations. Hence, we need to provide a way to control the write operations and prevent data loss.

Then, we saw that the unavailability of configuration servers can cause issues for the sharded cluster, and at least one configuration server should be available...