Disk space diagnosis

更新时间:
复制 MD 格式

The disk space diagnosis feature helps you locate the source of high disk usage when it cannot be identified directly. With a single click, the feature scans the disk of an instance and returns the usage distribution across categories and the top space consumers in each category, so that you can clean up the disk in time. The feature supports two scenarios: scanning kubelet disk usage on ACK or K8s nodes, and scanning system disk usage on ECS instances.

Limitations

  • Region limitations

    This feature is available in all regions except US (Virginia) and Saudi Arabia (Riyadh).

  • Operating system limitations

    Architecture

    Operating system

    x86 architecture

    • Rocky Linux 9.5

    • Rocky Linux 9.1

    • Ubuntu 20.04

    • Ubuntu 22.04

    • Ubuntu 24.04

    • Alibaba Cloud Linux 2

    • Alibaba Cloud Linux 3

    • Alibaba Cloud Linux 3 Pro

    • Alibaba Cloud Linux 4

    • Alibaba Cloud Linux 4 Pro

    • Alibaba Cloud Linux 3 Container-Optimized Edition

    • CentOS 7.6 and later

    • CentOS 8

    • Anolis OS 7

    • Anolis OS 8

    ARM architecture

    • Alibaba Cloud Linux 3

    • Alibaba Cloud Linux 3 Pro

    • Alibaba Cloud Linux 4

    • Alibaba Cloud Linux 4 Pro

Prerequisites

  • The following access policies are attached to your Resource Access Management (RAM) user:

    • AliyunECSReadOnlyAccess: read-only access to ECS.

    • AliyunSysomFullAccess: full access to SysOM.

  • Access to the operating system console is enabled.

Procedure

  1. Go to the operating system console.

  2. In the left-side navigation pane, choose System Diagnosis > Node Diagnosis.

  3. On the Node Diagnosis page, select the target instance, select Disk Analysis Diagnosis from the Diagnosis Type drop-down list, and click Start Diagnosis.

    Note
    • The scan takes about 1 to 2 minutes. A diagnosis report is generated automatically when the scan completes.

    • For an ACK or K8s node, the system scans usage of the kubelet data disk. For an ECS instance, the system scans usage of the system disk.

  4. After the scan completes, in the diagnosis records list, click View Report in the Actions column of the target record to view the report details.

Diagnosis report

The diagnosis report contains the following information:

Basic information

The basic information about the diagnosed instance, including the instance ID, instance name, and diagnosis time.

Diagnosis conclusion

An overall conclusion about disk usage based on the scan results.

Recommendations

Cleanup or optimization recommendations based on the diagnosis conclusion.

Diagnosis details

The diagnosis details consist of the following panels. The panels shown vary by scenario (ACK/K8s node or ECS instance).

Panel name

Description

Applicable scenario

Category Usage Distribution

A pie chart that shows the share of disk usage by category. For an ACK/K8s node, this includes seven categories: container image layers, Content Store, containerd metadata, container logs, system logs, kubelet data, and other directories. For an ECS instance, this includes two categories: system logs and other directories.

All

Top Container Image Layers

Displays the container image layers that consume the most space.

ACK/K8s node

Top Content Stores

Displays the Content Store entries that consume the most space.

ACK/K8s node

Top containerd Metadata

Displays the containerd metadata entries that consume the most space.

ACK/K8s node

Top Container Logs

Displays the container log files that consume the most space.

ACK/K8s node

Top System Logs

Displays the system log files that consume the most space.

All

Top kubelet Data

Displays the kubelet data directories that consume the most space.

ACK/K8s node

Top Other Directories

Displays directories that consume the most space but do not belong to the preceding categories.

All

Pod Usage Ranking

A table that ranks Pods by disk usage. Columns include Namespace, Name, Rootfs, Log, Ephemeral, and Total.

ACK/K8s node

Image Ranking

Displays the container images that consume the most space.

ACK/K8s node

Deleted but In-Use Files

Information about files that have been deleted but are still held by processes, including the original path (orig_path), file size (size), and holding processes (holders). These files have been deleted from the file system, but their disk space is not released because a process still holds their file descriptors.

All

FAQ

What do I do if du and df show different disk usage?

The du command counts the size of files that actually exist in the file system, whereas the df command counts the used blocks of the file system. When a file has been deleted but is still held by a process, du does not include it, but df still counts its blocks as used, which causes the two values to differ. You can find such files in the Deleted but In-Use Files panel of the diagnosis report and restart the holding processes to release the space.

Why doesn't the diagnosis report for an ECS instance show the container-related panels?

The Top Container Image Layers, Top Content Stores, Top containerd Metadata, Top Container Logs, Top kubelet Data, Pod Usage Ranking, and Image Ranking panels are shown only for ACK/K8s nodes. The ECS instance scenario scans only the system disk and does not include container runtime data, so these panels are not shown.

Does the feature support scanning data disks?

Currently, the disk analysis diagnosis feature scans only the system disk for ECS instances and only the kubelet data disk for ACK/K8s nodes. Scanning other data disks is not supported.