EC2/EBS Storage I/O Diagnostic Troubleshooter (Part 2)
Use the interactive troubleshooter below to identify your EC2/EBS storage I/O symptom, review the raw evidence, understand the root cause, and apply the recommended fix.
🚨 Step 1: What specific error symptom are you experiencing?
Please click the most accurate description:
Quick Reference Table
| # | Scenario | Key Error Signal | Root Cause | The Fix |
|---|---|---|---|---|
| 5 | latency — SSD write amplification and dramatically reduced I/O performance | "This pattern results in significantly increased write amplification, increased latency, and dramatically reduced I/O performance." | Writing unaligned bytes or small block sizes to SSD instance store volumes filled the disk, forcing the controller to execute expensive garbage collection. | Leave 10% of the volume unpartitioned for over-provisioning or configure the OS to use the TRIM command. |
| 6 | abort — NVMe Abort issued due to exceeded I/O timeout threshold | "The Abort command is an NVMe Admin command that is issued to abort a specific command... typically issued by the device driver to storage devices that have exceeded the I/O operation timeout threshold." | The Linux device driver issued an NVMe Abort because the storage device exceeded the configured I/O timeout threshold. | Increase nvme_core.io_timeout to 4294967295 or use instance types that natively support Abort (e.g., R5b, M6i). |
| 7 | EBS / BurstBalance — I/O credit bucket depleted on gp2 volume | "When I/O demand is greater than baseline performance, the volume spends I/O credits... You can monitor the I/O credit balance for a volume using the Amazon EBS BurstBalance metric in Amazon CloudWatch." | The workload's I/O demand exceeded the gp2 volume's baseline limit, entirely depleting its 5.4 million I/O credit bucket. | Modify the volume to gp3 or increase its size using aws ec2 modify-volume. |
| 8 | volume — Instance boots from wrong volume due to duplicate partition label | [ec2-user ~]$ sudo e2label /dev/xvda1 / / [ec2-user ~]$ sudo e2label /dev/xvdf1 / | The initial ramdisk booted from the wrong attached volume because multiple volumes concurrently possessed the identical / partition label or UUID. | sudo e2label /dev/xvdf1 old/ (ext4) or sudo xfs_admin -L old/ /dev/xvdf1 (xfs). |