Skip to main content

EFS Hands-On

Key Takeaways​

The Core Lifecycle Setup​

  • The Restorative Logic: Stephane demonstrates setting up policies to transition files into Infrequent Access (IA) or Archive tiers based on age (e.g., 30 or 90 days of no activity). Critically, you can set an option to transition files back to Standard on first access if you expect them to be heavily reused once woken up.
    EFS Lifecycle Management
  • Throughput Decoupling: Stephane clarifies the layout of the three modes: Bursting (scales with storage size), Provisioned (you pay to lock in a specific baseline speed), and Elastic (the modern recommended default that scales instantly from 0 to hundreds MB/s - perfect for unpredictable I/O). EFS Throughput Modes

The Security Group Setup​

  • The inbound rule: For your EC2 instance to talk to EFS, the EFS mount targets must allow NFS traffic on port 2049.
  • Security Group referencing: instead of whitelisting specific IP blocks, AWS automatically creates an EFS Security Group (efs-sg-x) that whitelist the EC2 instance's security group ID as the source. EFS automatic security group referencing

Automated bootstrapping via EC2 launch​

  • In the past, you had to log into Linux and type out mount commands manually.
  • Now, you simply assign your subnet, check the "Automatically mount EFS file system" box in the EC2 wizard and define a mount point (like /mnt/efs/fs1).
  • AWS handles the heavy lifting by injecting the exact required mount scripts directly into the instance's User Data automatically. EFS Automatic Mounting
    User Data Script with EFS Mounting

Cross-AZ Live Verification​

We can proves that EFS is truly multi-AZ by launching two EC2 instances in different AZs, mounting the same EFS file system on both.

  • Instance A: Creates a new instance called instance-a, In the network settings, select the Subnet that belongs to ap-southeast-2a. Attach the EFS file system that you have created before and specify the mount point (e.g., /mnt/efs/fs1).
    • After the instance is up, SSH into it and create a new file
    echo "hello world" > /mnt/efs/fs1/hello.txt
  • Instance B: Create another instance called instance-b, but this time select the Subnet that belongs to ap-southeast-2b. Attach the same EFS file system and specify the same mount point.
    • After the instance is up, SSH into it and check if the file created by instance-a is visible in instance-b
    cat /mnt/efs/fs1/hello.txt
  • The Takeaway: This confirms true concurrent multi-AZ reading and writing over a POSIX-compliant file system.

Clean-up​

Stephane reminds us to clean up resources to avoid stray costs:

  1. Termnate both EC2 instances (instance-a and instance-b).
  2. Delete the EFS file system
  3. Delete any security groups that were automatically created for EFS (e.g., efs-sg-x and instance-sg-x)