Ravindra BagaleCourses & study guides मराठी Track your progress

Chapter 17: Scaling

17.8 CPU, RAM, or storage

By Last updated:

Why "make it bigger" is three different jobs

CPU calculates. RAM holds data while the process runs. Storage is the disk, which on EC2 is the EBS volume, not the instance type. t3.micro to t3.small adds memory (about 1 GiB to about 2 GiB). nproc stays 2. The disk does not grow.

What you measure, and what you change

You see Resource Change this Do not change this
CPU about 90%, free -h still has memory free, pages slow CPU Another instance, or a type that adds vCPUs A bigger disk. t3.small, which still has 2 vCPUs
free -h nearly empty, swap in use, many PHP or Node workers, a large MySQL buffer, or the kernel kills a process (OOM) RAM More memory (t3.small), or fewer workers A larger EBS volume
df -h says 100%. Logs, uploads, or database files cannot be written. CPU can be low Storage A larger EBS volume, then grow the filesystem (Chapter 13) The instance type

Choose CPU, RAM, or disk from what you measured

How much to add. One step, not a jump. CPU: add 1 instance (desired 1 to 2), or pick the next type that actually increases nproc. Do not jump from t3.micro to a huge type. RAM: go from about 1 GiB (t3.micro) to about 2 GiB (t3.small). Disk: if the volume is 8 GiB and df -h says full, grow it by a few GiB (Chapter 13 uses 8 to 12), not to a giant disk.

CPU, step by step.

  1. Run nproc. Write the number.
  2. Run free -h. You should still see free memory.
  3. Open CloudWatch. You should see CPUUtilization near 90.
  4. Do not change the instance type to t3.small for this row. That does not add a vCPU.
  5. Add one more instance behind the load balancer, as in section 17.3.
  6. You should see two servers, each nearer 45% than 90%.

RAM, step by step.

  1. Run free -h.
  2. You should see almost no available memory, or swap in use.
  3. Stop the instance.
  4. Change t3.micro to t3.small only. That is the one-step increase, about 1 GiB to about 2 GiB.
  5. Start the instance.
  6. Run free -h again. You should see about 2 GiB.
  7. This does not fix a disk that df -h says is full.

Disk, step by step.

  1. Run df -h.
  2. You should see / at 100% and 0 available.
  3. Do not change the instance type. The same full volume would come back.
  4. In EC2, select the volume.
  5. Choose Actions, then Modify volume.
  6. Increase the size by a few GiB, the way Chapter 13 does (for example 8 GiB to 12 GiB).
  7. On the instance, run sudo growpart on the right device, as in that chapter.
  8. For XFS run sudo xfs_growfs -d /. For ext4 run sudo resize2fs on the partition.
  9. Run df -h again. You should see free space.

Lab

Chala.

  1. Run nproc.
  2. Run free -h.
  3. Run df -h.
  4. If CPU is 90 and memory is free, write "add one instance."
  5. If swap is full and CPU is low, write "change to t3.small, about 2 GiB."
  6. If the filesystem is 100%, write "grow the EBS volume a few GiB. Do not change the type."