CCE AI Assistant

CCE AI assistant provides intelligent Q&A to address your inquiries regarding Huawei Cloud services. Through the chat window, you can ask questions anytime and anywhere, enabling efficient understanding of Huawei Cloud. Integrating knowledge from hundreds of Huawei Cloud services and public resources available online, the assistant leverages foundation models to accurately identify user intention and respond using natural language.

The typical problem types include:

• Concept analysis: The assistant interacts with you using natural language to deliver in-depth explanations of technical terms, product functions, and operations within Huawei Cloud services, such as compute, storage, network, and database services. You can quickly obtain structured knowledge. Using Huawei Cloud documentation, best practices, and foundation models, it automatically identifies user intention and provides hierarchical explanations, covering basic concepts, advanced usage, and configuration examples; and it dynamically adjusts explanation complexity to adapt to your technical background. For example, when discussing CCE, it can link concepts like Kubernetes and microservices, offering guidance on deployment procedures or troubleshooting.

  Example prompts:

  • What are the differences between Elastic Cloud Server (ECS) and Bare Metal Server (BMS)?
  • What is a multipart upload in Object Storage Service (OBS)?
• Purchase consultation: The assistant uses intelligent recommendation algorithms and dynamic pricing models to help you select cloud products, optimize costs, and provide transaction support for services in all categories such as compute, storage, and AI. You can ask questions using natural language. For example, you could ask, "How do I select a cloud server package suitable for a small- or medium-sized enterprise?" or "What discounts can I enjoy when I purchase OBS?" It accurately interprets these questions and provides precise advice tailored to your needs. By analyzing user requirements for performance, budget, and scenario, the assistant delivers precise suggestions, such as detailed product comparisons, configuration recommendations, cost estimates (evaluating pay-per-use versus yearly/monthly plans), insights into current promotions, and an explanation of relevant contract terms. For example, if you inquire, "How can I purchase an ECS and get a discount?" the AI assistant can recommend ECS flavors tailored to your needs, calculate the associated fees, and showcase current promotional offers, such as discounts for new users. Additionally, it provides step-by-step guidance to make it easier for you to complete yourself purchase. The assistant supports multi-round conversations, for example, "How can I deploy an HA e-commerce system?" "How can the system be set up with a limited budget?" "How can I design the optimal solution?" to address complex requirements. Additionally, it synchronizes the latest product details and pricing policies directly from Huawei Cloud KooGallery in real time.

  Example prompts:

  • Is yearly/monthly Relational Database Service (RDS) more cost-effective than pay-per-use?
  • How can I apply for a batch purchase discount as an enterprise user?
• Usage consultation: The assistant focuses on resolving specific user problems and optimization needs across scenarios like compute instances, databases, and containers. You can ask questions like, "What can I do if an ECS cannot be accessed?" or "How do I optimize RDS database performance?" The assistant interprets these queries and delivers guidance, including detailed steps for operations, fault diagnosis recommendations, and strategies for enhancing performance. It leverages real-time monitoring data, log analysis, and intelligent diagnostic algorithms to deliver tailored solutions. For example, it automatically detects incorrect network configurations in instances and provides step-by-step rectification guidance, and it analyzes database loads to recommend optimized indexing strategies for improved performance. Multi-round interaction allows the assistant to tackle complex issues, such as configuring cross-region disaster recovery. The assistant also integrates with the console and with APIs to directly showcase resource statuses, create configuration templates, or redirect you to relevant operation pages, enhancing efficiency in problem resolution.

  Example prompts:

  • What can I do if the OBS access latency is high?
  • How do I configure a load balancer to distribute traffic?

Huawei Cloud intelligent assistant provides cluster health checks, allowing you to monitor and optimize cluster performance through natural language instructions and scheduled inspection settings. The key health checks it performs include:

• Load checks: The assistant monitors CPU, memory, and disk I/O usage for each cluster node. It identifies overloaded or idle nodes and suggests load balancing strategies.

  Example prompts:

  • What should I do if the CPU usage of my cluster is too high?
  • How do I check the load balancing status of each node?
• Configuration checks: The assistant evaluates cluster settings like replicas, ReplicaSets, and auto scaling policies for risks or improvement opportunities.

  Example prompts:

  • Is the replica configuration of the current cluster sufficient?
  • How do I adjust the threshold for auto scaling?
• Security checks: The assistant scans for security vulnerabilities, such as unfixed patches and open high-risk ports within clusters and checks whether protection measures, such as firewalls and intrusion detection systems, are active and effective.

  Example prompts:

  • Does the cluster have any unfixed security vulnerabilities?
  • How do I disable unnecessary open ports?
• Network checks: The assistant checks the cluster network connectivity, bandwidth utilization, VPC configuration, and sees if there is excessive cross-node communication latency. These measures help ensure network stability and efficiency.

  Example prompts:

  • What do I do if the network latency between cluster nodes is too high?
  • How do I optimize the network bandwidth across AZs?
• Storage checks: The assistant monitors storage space usage, assesses I/O performance, and verifies data backup status. It identifies potential issues such as insufficient storage capacity or performance bottlenecks.

  Example prompts:

  • How do I expand the OBS capacity when storage space starts to run out?
  • How do I improve the read and write speed of database disks?
• Lifecycle checks: The assistant traces the run times, expiration times, and automatic destruction policies of cluster instances to prevent resource expiration or unexpected interruptions.

  Example prompts:

  • How long will my CCI instances last?
  • How do I configure when a cluster instance is automatically destroyed?
• Security and permission configuration: The assistant verifies that user permissions assignment (such as IAM roles) and access control policies (such as security group rules) comply with the principle of least privilege to prevent unauthorized access.

  Example prompts:

  • Does this user have excessive permissions?
  • How do I prevent specific IP addresses from accessing a cluster?
• Scheduled inspections: The assistant allows you to configure periodic health checks for clusters using natural language instructions, for example, "Check cluster health at 08:00 every day." Reports are generated automatically, and alarms are reported to notify you of potential issues.

  Example prompts:

  • Set the cluster security status to be automatically checked every Friday.
  • Set the storage space usage to be checked at 02:00 every day.

Huawei Cloud CCE AI assistant offers advanced multi-dimensional resource query and statistics analysis capabilities. You can use natural language instructions to quickly locate resources or analyze usage across core services such as ECS, CCE, VPC, Elastic Load Balance (ELB), Elastic Volume Service (EVS), RDS, Distributed Cache Service (DCS), Document Database Service (DDS), and OBS. The core capabilities include:

• General query: You can quickly locate specific resources or generate usage reports based on parameters such as resource ID, name, region, tags, enterprise projects, and time ranges.
• Extended/Associated query: You can perform cross-service queries based on resource associations (for example, obtaining all ECSs and EVS disks attached to them in a VPC) or specific service scenarios, addressing complex requirements (for example, obtaining unused load balancers in an enterprise project).

Built using powerful Pangu models and the DeepSeek inference engine, Huawei Cloud CCE AI assistant achieves deep integration of Huawei Cloud resource metrics, logs, and event data, including Prometheus monitoring metrics, container events, and container logs. This integration enables it to associate real-time metrics such as CPU and memory usage, critical events like pod eviction or OOMKilled, and patterns of exceptions in logs through natural language. CCE AI assistant's resource metric log query is designed to streamline the identification of resource issues and enhance operational efficiency. By reducing manual query efforts, it significantly accelerates fault resolution. Its advanced analytical features automatically detect sudden changes in metrics, log anomalies, and recurring events, providing precise root cause identification for problems like memory leaks and insufficient resources. Moreover, it supports cross-dimensional decision-making by using multi-dimensional aggregation analysis and presents results through intuitive visualizations. This feature serves as a foundation for resource optimization and capacity planning, offering actionable insights such as adjustments to pod resource quotas. These capabilities empower you to make informed decisions while boosting the overall accuracy and effectiveness of O&M processes. The core capabilities include:

• Prometheus metric queries
  • You can obtain real-time and historical metrics (such as CPU usage, memory usage, network traffic, and pod restart times) for the pods, containers, and nodes in a CCE cluster. Natural language queries are converted into PromQL syntax for complex queries. Example: "Obtain the average CPU usage of pods whose names contain web in the last hour."
  • Prometheus alarms and events are automatically associated, and correlations between metric anomalies and alarms are analyzed. Example: "Tell me how many pods have triggered memory alarms."
• Container event analysis
  • The location of events can be quickly identified based on time ranges, event types (statuses such as Warning or Normal), and causes (such as evicted or OOMKilled). Example: "Obtain the number of OOMKilled events in the cluster in the last 24 hours."
  • Event causes are inferred based on event descriptions and contextual data like pod names, node IP addresses, and timestamps. Example: "Identify possible causes of pod evictions."
• Container log search
  • Logs can be obtained based on keywords, regular expressions, time ranges, and resource associations (such as pod names, container IDs, and namespaces).
  • Key fields (such as error codes, request IDs, and timestamps) are automatically extracted from logs and structured reports are generated. You can also take advantage of correlation analysis with Prometheus metrics or events. Example: "Analyze correlations between pod log errors and increased CPU usage."
• Smart correlation analysis
  • Cross-resource association queries: Multi-dimensional diagnosis is conducted based on relationships between resources. Example: "Obtain the CPU usage of all ECSs in a VPC and the IOPS of the EVS disks attached to these ECSs."
  • Abnormal pattern identification: Sudden changes in metrics, abnormal log patterns, and frequent event triggering can be automatically detected based on AI models. Example: "Detect abnormal fluctuations in pod memory usage and generate reports."

Huawei Cloud CCE AI assistant uses natural language to help you manage Kubernetes cluster resources, such as pods, Deployments, and Services. It facilitates tasks like creating, deleting, and updating resources. For security reasons, the assistant does not execute commands directly. Instead, it generates kubectl commands or YAML configuration files based on your requirements. After your confirmation, the commands or files can be copied directly into the Huawei Cloud console or terminal for execution. The following shows some examples of the usage of this feature:

• Resource creation: Generate a Deployment YAML file to define application replicas.
• Configuration updates: Generate rolling update commands to modify pod resource quotas.
• Resource deletion: Generate the kubectl commands to delete unnecessary Services.

In addition, the assistant can help you access a pod to run commands.

Powered by the Pangu models and DeepSeek thinking model, Huawei Cloud CCE AI assistant leverages a multi-expert agent architecture based on the expert experience knowledge base to intelligently explore, detect, and diagnose cluster issues. It offers six core diagnostic capabilities through natural language interaction: node diagnosis, pod diagnosis, service diagnosis, ingress diagnosis, memory diagnosis, and network diagnosis. These features help you swiftly identify root causes of faults and deliver professional solutions. You can call the Huawei Cloud AI assistant to diagnose resources in two ways. First, click the AI assistant icon located in the sidebar. Second, hover over an abnormal resource and click the option in the pop-up message.

• Node diagnosis: The assistant examines node resource usage, system logs, and network status to identify node anomalies. These include nodes in the ContainerCreating or Unknown state, ECS status issues, kubelet component failures, runtime errors, full disk storage, and high CPU loads.

  Example prompts:

  • Scenario: The node is abnormal, for example, the disk may be full.

    Prompt: I received an alarm indicating that the disk usage of node-01 exceeds 95%. What actions should I take?

  • Scenario: The CPU load on a node is too high.

    Prompt: How can I verify whether the CPU load on node-01 is abnormal?

• Pod diagnosis: The assistant analyzes the pod lifecycle status, resource usage, and container run logs to identify various issues. These include pods that have been scheduled but not processed by kubelet, delays caused by waiting for storage volumes to be ready, and evictions due to disk space, disk inodes, or insufficient memory. It also diagnoses sandbox container creation failures, pods stuck in a Terminating state, and container OOM exceptions. Additional issues include abnormal exits, pods in a CrashLoopBackOff state, problems with the ContainerCreating state, and image pulling errors or timeouts.

  Example prompts:

  • Scenario: A pod is evicted due to insufficient node space.

    Prompt: Why was the cache-worker-456 pod evicted?

  • Scenario: A pod image pull times out.

    Prompt: What can I do if the action timed out when I tried to pull an image for the web-app-789 pod?

• Service diagnosis: The assistant verifies the consistency between the Service and the load balancer configuration, including billing mode and certificate alignment. It evaluates the health status and quota limits, such as backend server quota, listener quota, and associated load balancer quota, of the backend pods. Additionally, it detects service exceptions or cases where the load balancer is missing.

  Example prompts:

  • Scenario: Checking the configuration of a Service

    Prompt: Help me diagnose the load imbalance of the backend pods of the api-service Service.

• Ingress diagnosis: The assistant checks for ingress configuration conflicts, TLS configuration, and health check failures.

  Example prompts:

  • Scenario: There is a conflict with the ingress configuration.

    Prompt: Why is not the base-url-scheme configured for the api-ingress ingress working?

• Memory diagnosis: The assistant detects memory leaks, analyzes memory usage, residual memcg data, fragmentation, and THP wastage, and pinpoints memory resource anomalies at the cluster or pod level.

  Example prompts:

  • Scenario: There is a memory leak.

    Prompt: Check whether there is any memory leak in the memory-heavy-01 pod.

• Network Diagnosis: The assistant examines the connectivity between pods or nodes, the status of links connecting pods or nodes to Services or DNS, interruptions in public network access links, and connection issues with the LoadBalancer Services.

  Example prompts:

  • Scenario: There is an abnormal network connectivity between pods.

    Prompt: Check the network connectivity between the frontend-123 pod and the node-09 node.

  • Scenario: There is a DNS resolution failure.

    Prompt: Check why DNS resolution failed for the web-app-789 pod.

Huawei Cloud CCE AI assistant enables full-lifecycle service ticket management through natural language interaction. Its capabilities include ticket submission, status queries, withdrawal, modification, and accessing historical records. The capabilities include:

• Submitting a service ticket: You can describe a problem using natural language, such as insufficient node disk space or an abnormal pod exit. The AI assistant automatically identifies the problem type, links it to a diagnosis (see Intelligent Diagnosis of Cluster Resource Faults), and generates a standardized service ticket. You can attach logs, screenshots, or diagnosis reports to the ticket.

  Example: "Submit a service ticket based on the diagnosis records of the web-app-789 pod."

• Ticket status queries: The assistant easily retrieves the current progress, assigned handler, and expected resolution time of a ticket based on its ID, keywords (like disk scale-out), time ranges (such as last week), or its status (such as processing).

  Example: "List the service tickets that are submitted but not processed."

• Withdrawing a service ticket: The assistant can withdraw a ticket that has not yet been processed, whether it was submitted by mistake or the problem has been resolved.

  Example: "Please withdraw the service ticket related to api-service."

• Ticket modification: The assistant can update or refine key details of a submitted ticket, such as the problem description or its priority.

When asking questions, include specific details such as the names, IDs, regions, and tags of the resources in questions. This helps the AI assistant identify faults more precisely. For example, if you encounter an ECS connection issue, providing the ECS ID and region enables the assistant to diagnose the problem more quickly.

Background

Share relevant background information, including the environment, the time, and any relevant operations performed. This allows the AI assistant to fully understand the context and deliver precise solutions. For example, if a pod is restarting frequently, describe when the issue occurs and provide operation logs to help the AI assistant pinpoint the root cause.

Error Logs

If the issue involves error logs or exception details, share this information with the assistant to speed up the diagnosis. For example, if a pod fails to start, key details from the error logs can help the assistant diagnose the issue more precisely.

Technical Prompts

Use precise technical terms when asking questions to improve the accuracy and speed of the AI assistant's response. For example, instead of asking, "Why cannot I access my ECS?" you could say, "Why cannot I access an ECS using SSH?"

Brief, Concise Prompts

Keep your problem statement straightforward and clear. Avoid overly complex or lengthy explanations. This ensures the AI assistant quickly grasps the issue and provides a solution. For example, instead of asking, "What should I do if I receive an alarm indicating that the CPU loads of node-01 exceed a configured threshold and services may be affected?" you could say, "What can I do if the CPU loads of node-01 are too high?"

Narrowing the Problem Scope

When a problem is large or complex, it is important to narrow its scope and provide more details. The AI assistant can deliver more focused solutions based on your feedback. If the problem is not resolved on the first attempt, try adding more background information or a more detailed description of the steps.

Escalating the Problem

For complex problems or situations requiring further support, you can instruct the AI assistant to generate a service ticket or offer a more comprehensive solution. If the problem persists after multiple attempts, you can instruct the assistant to submit a service ticket for further processing.