Mongodb crashes due to too many open files error

Problem 

When running MongoDB in production, configuring the open file descriptor limit (nofile) is critical. Ubuntu's default limit (1024) is insufficient for production workloads and may result in connection failures or 'Too many open files' errors under high load.

Wherever we are running production workloads like AWS EC2 , Azure VM , GCP Cloud and DigitalOcean it will be same problem 

Why Open File Limits Matter MongoDB uses file descriptors for :

• Client connections
• WiredTiger data files
• Journal files
• Log files
• Internal sockets

Each active connection and file consumes a descriptor. High-traffic environments require a significantly higher limit.

Recommended Production Value Set the open file limit to 1048576

This is typically the maximum allowed by the Ubuntu kernel and is effectively unlimited for most real-world workloads.

Configuration Steps (systemd)

Create or edit the systemd override file:

sudo systemctl edit mongod or vi /etc/systemd/system/mongod.service.d/override.conf

Add the following configuration:

[Service]
LimitNOFILE=1048576

Reload and restart MongoDB:

   sudo systemctl daemon-reexec
   sudo systemctl daemon-reload
   sudo systemctl restart mongod

Also Change limits configuration ,

vi /etc/security/limits.conf

mongodb soft nofile unlimited

mongodb hard nofile unlimited

root soft nofile unlimited

root hard nofile unlimited

Verification 

cat /proc/$(pidof mongod)/limits | grep "open files"

Expected Output 

Kernel Limitation 

Linux does not support true unlimited file descriptors. Even when setting LimitNOFILE=infinity, the value is capped by the kernel parameter fs.nr_open.On Ubuntu, this is commonly 1048576.

Final Recommendation 

For MongoDB production environments ,

• Set LimitNOFILE to 640000 or 1048576
• Ensure fs.nr_open and fs.file-max are properly configured
• Always verify after restart

Proper OS tuning ensures stability, prevents resource exhaustion, and maintains optimal MongoDB performance under high concurrency.