Operating system console release notes

更新时间:
复制 MD 格式

This article details the latest updates to the operating system console. These enhancements improve system management to support your evolving applications, resulting in better operating system performance, increased efficiency, and an improved user experience.

Background information

  • The operating system console supports specific versions of CentOS, Alibaba Cloud Linux, Anolis, and Ubuntu.

  • See the relevant documentation for applicable regions.

May 2026

Feature

Module

Description

Release date

References

Component management

SysOM

Adds management and diagnostics for Alibaba Cloud Linux 4.

2026-05-30

Component management

System diagnosis

One-click diagnosis

1. Adds diagnostics for critical system service exceptions.

2. Adds diagnostics for high pod load.

3. Adds one-click diagnostics for Lingjun.

4. Adds multi-dimensional GPU performance diagnostics for DLC training jobs.

5. Adds CPU load performance diagnostics for DLC training jobs.

6. Adds diagnostics for EAS job snapshot exceptions (checkpoint and restore).

7. Adds pod memory status checks and feature configuration diagnostics.

2026-05-30

One-click diagnosis

Memory diagnosis

Adds huge pages analysis and Java flame graph analysis.

2026-05-30

Java memory diagnosis

Scenario-based diagnosis

Adds disk usage analysis for nodes and containers, and file analysis for specific directories.

2026-05-30

Disk space diagnosis

Agent monitoring

Token analysis

Adds agent token usage statistics and analysis of traces and session interruptions.

2026-05-30

Agent monitoring

April 2026

Feature name

Feature module

Description

Release date

Related documents

GPU Performance and Diagnosis

GPU resource topology

Optimizes pod labels and enhances overall functionality.

2026-04-30

GPU resource topology

component management

configuration management

Adds an "Auto" memory pressure option for FastOOM and supports kernel-level QoS capability.

2026-04-30

SysOM FastOOM Feature Guide

OS Copilot

Enhances O&M capability with support for Skills.

2026-04-30

SysOM Diagnostic Skill Guide

March 2026

Feature name

Feature module

Description

Release date

Documentation

System health and diagnostics

System diagnostics

The memory diagnostics feature now analyzes Go application memory and system memory, and provides an optimized display for Glibc memory.

2026-03-30

Go application memory diagnostics

Improved one-click diagnosis performance and added support for CentOS versions earlier than 7.6.

2026-03-30

One-click diagnosis

System health

Added dentry anomaly detection for memory resources.

2026-03-30

System health

Component management

Component management

Enhanced error code messages and improved component installation performance.

2026-03-30

Component management

January 2026

Feature name

Feature module

Description

Release date

References

OS Copilot

intelligent agent

Adds high CPU load diagnosis, which automatically identifies and analyzes spikes in CPU utilization to pinpoint hot threads or methods.

2026-01-30

OS Copilot user guide

system diagnosis

Memory Panoramic Analysis

Adds output support for glibc local memory allocation to provide a more comprehensive analysis of native memory usage.

2026-01-30

Memory Panoramic Analysis

System Health and Diagnosis

System Health

Adds intelligent diagnosis for abnormal alarms, using a large model to provide context-aware root cause analysis and fix suggestions.

2026-01-30

System Health

December 2025

Feature name

Feature module

Description

Release date

Related documents

GPU performance and diagnostics

GPU resource topology

Provides two unified views of GPU clusters to view GPU resource usage and analyze collective communication for distributed training.

  • Resource topology view: Shows the busy and idle status of GPU resources in the cluster and provides details on resource usage by jobs.

  • Cluster communication view: Analyzes the collective communication topology, communication data flow, and communication exceptions for tasks.

2025-12-31

GPU resource topology

GPU diagnostics

Adds slow node diagnostics for large model training to help you quickly identify the nodes responsible for training slowdowns.

2025-12-31

GPU diagnostics

System diagnostics

memory panoramic analysis

The memory panoramic analysis for Java memory diagnostics now supports large file uploads and diagnostics for unmanaged instances. It also adds on-heap/off-heap flame graphs and provides statistics on memory changes during memory profiling.

2025-12-31

memory panoramic analysis

OS Copilot

intelligent diagnostics

OS Copilot can now diagnose the MCP service. Third-party intelligent agents can use the console's diagnostic features through MCP to diagnose issues such as high memory usage, high CPU utilization, and performance jitter.

2025-12-31

OS Copilot user guide

November 2025

Feature name

Feature module

Overview

Release date

References

component management

configuration management

Adds support for displaying regular expression match information in FastOOM.

2025-11-30

SysOM FastOOM User Guide

Adds zombie Memcg reclaim.

2025-11-30

Zombie Memcg Reclaim User Guide

system diagnosis

one-click diagnosis

Enables one-click scanning for common OS issues.

2025-11-30

One-Click Diagnosis

memory overview

Adds Java memory diagnostic analysis to memory overview.

2025-11-30

Memory Overview Analysis

OOM diagnosis

Enhances OOM data collection in SysOM.

2025-11-30

SysOM EnhanceOOMCollector

historical diagnosis

Introduces historical diagnosis to analyze past OOM events on ACS and ECI instances.

2025-11-30

Historical Diagnosis

OS Copilot

Intelligent Memory Diagnosis

When you ask the intelligent assistant about a related issue, it guides you to collect information, then automatically runs memory overview, Java memory diagnosis, and OOM diagnosis to generate a diagnostic report with relevant links.

2025-11-30

OS Copilot User Guide

October 2025

Feature name

Feature module

Description

Release date

Related documentation

Component management

Component management

Adds support for Rocky Linux 8.8 and later.

2025-10-30

Component management

Abnormal event push

Abnormal event alert

Event content now includes k8s label information and GPU exception information.

Downtime alerts now include diagnostic information.

2025-10-30

Abnormal event alert

AI performance diagnosis

AI flame graph

Introduces the AI flame graph feature, enabling performance profiling for AI applications.

2025-10-30

Continuous GPU profiling

Homepage overview

System overview

A new overview page in the console includes the following sections:

  • Status of your managed resources

  • Distribution of abnormal events

  • Subscription status

  • Quick start guide

  • Latest updates for the console

2025-10-30

System health

August 2025

Feature name

Feature module

Description

Release date

Documentation

Java memory diagnostics

System diagnostics

Provides memory analysis for Java applications. It scans the memory usage of a Java process and its containing pod, detailing the memory footprint from both JVM and application-level perspectives.

2025-08-30

Java memory diagnostics

Pod diagnostics

System diagnostics

Adds a diagnostic mode to support memory diagnostics for pods on ACK and ACS.

2025-08-30

Pod diagnostics

Process hotspot tracking

System observation

Lets you query hotspot history.

2025-08-30

Process hotspot tracking

Abnormal event push

Abnormal event alerts

Adds support for pushing downtime events.

2025-08-30

Abnormal event alerts

July 2025

Feature name

Feature module

Description

Release date

Related documentation

FastOOM

Component management

When enabled, this feature monitors node-level memory pressure and triggers FastOOM to kill processes, preventing system slowdowns or unresponsiveness.

2025-07-30

SysOM FastOOM Feature Guide

Pod diagnosis

System diagnosis

Diagnoses pods in ACK clusters. You can specify a pod for diagnosis by its Kubernetes namespace and pod name.

2025-07-30

Pod diagnosis

Alert delivery

System overview

Sends multi-channel notifications for exception events like abnormal CPU utilization, enabling rapid issue response and resolution.

2025-07-30

Exception event alerts

Memory overview

System diagnosis

Added tcp_mem and socket leak detection.

2025-07-30

Memory overview analysis

Crash diagnosis

System diagnosis

Adds support for hardlock and hung task analysis.

2025-07-30

Crash diagnosis

AI Profiling

AI Profiling

Lets you export diagnostic reports for offline viewing.

2025-07-30

AI Profiling

Now supports Ubuntu 22.04 and Ubuntu 24.04.

June 2025

Feature name

Feature module

Description

Release date

Related documentation

Abnormal health event

System overview

Added checks for tcp_mem, socket leaks, and I/O read/write latency exceptions.

2025-06-30

System health

Configuration management

Component management

Supports configuring FastOOM to terminate processes based on flexible configurations, preventing system slowdowns or unresponsiveness before an OOM.

2025-06-30

SysOM FastOOM user guide

Canary deployment

Component management

Supports canary deployment of components in an ACK cluster by node count, percentage, or ACK node labels.

2025-06-30

Canary deploy components in an ACK cluster

Diagnosis for unmanaged instances

System diagnosis

You can use system diagnosis directly on unmanaged instances.

2025-06-30

Use the operating system console

AI observability

AI observability

  • Supports the iteration collection mode.

  • Supports a statistical chart for GPU kernel function call duration (μs).

  • Supports data collection from PyTorch, FLOPs, GPU memory, and the Python stack.

  • Supports AI differential analysis.

2025-06-30

AI profiling

March 2025

Feature name

Feature module

Overview

Release date

References

node monitoring

system management

View metrics for network, I/O, scheduling, and memory on your nodes.

2025-03-31

system management

AI diagnosis

AI diagnosis

Diagnose system exceptions in AI workloads and generate findings, GPU status, and AI job status.

2025-03-26

GPU diagnosis

Java hotspot tracking

process hotspot tracking

Track hotspots in a Java process and collect Java stack information.

2025-03-12

process hotspot tracking

December 2024

Feature name

Feature module

Overview

Release date

References

Cluster health

system overview

View the health status of multiple clusters and switch between them.

2024-12-30

cluster health

Anomaly feedback

Provide feedback on anomalies to adjust the reporting frequency and baselines.

2024-12-30

One-click IO diagnosis

system diagnosis

Diagnoses common issues such as high IO latency, IO bursts, and IO wait.

2024-12-30

one-click IO diagnosis

Network jitter diagnosis

Analyzes network jitter caused by slow packet reception, soft interrupts, and scheduler issues, providing specific causes and recommended solutions.

2024-12-30

network jitter

System load diagnosis

Analyzes the causes and details of anomalies in the one-minute average system load (load1 metric).

2024-12-30

system load diagnosis

Downtime diagnosis

Analyzes the causes of downtime and provides recommended fixes.

2024-12-30

downtime diagnosis

AI infra observation

system observation

AI infra observation uses dynamic injection for GPU profiling to analyze operators, scheduling, and kernel functions during AI job execution. This enables non-intrusive, highly flexible, and low-overhead observation of AI jobs.

2024-12-30

AI profiling

Cluster deployment

component management

Component management now supports managing ACK clusters and enables auto scaling for instances within the cluster.

2024-12-30

component management

Unsubscribe

subscription management

You can now use the console to unsubscribe from CentOS security updates and the Alinux 2 ELS subscription.

2024-12-30

subscription management

Diagnostic guidance

OS Copilot

Provides diagnostic guidance for anomalies such as OOM.

2024-12-30

OS Copilot

Documentation links

OS Copilot answers now include documentation links.

2024-12-30