Updated on 2026-11-20 GMT+08:00

Overview

You can flexibly create alarm rules on the Cloud Eye console. You can create an alarm rule for a specific metric or event or use the alarm template to create alarm rules for multiple cloud service resources in batches.

Cloud Eye provides you with default alarm templates tailored to each service. In addition, you can also create custom alarm templates by modifying the default alarm template or by specifying every required field.

You can enable Alarm Notifications when creating alarm rules. When a metric reaches the threshold specified in an alarm rule, Cloud Eye will notify you by SMS, email, HTTP or HTTPS message, FunctionGraph (function), FunctionGraph (workflow), WeCom, DingTalk, Lark, or WeLink Chatbot in real time. You can track the service status and establish programs accordingly to handle the alarms.

Table 1 Applicable scenarios

Alarm Type

Scenario

Metric

Cloud Eye can collect metric data of cloud services, such as CPU usage and memory usage. You can monitor metrics to track the status of cloud services.

You can set alarm rules and notifications for core service metrics. When a metric triggers the preset threshold, Cloud Eye sends you a notification. This helps you identify and handle abnormal monitoring data in real time.

Event

Events are key operations or statuses of a cloud service, for example, restarting a VM.

You can set event alarm rules for key business events or cloud resource operations. When a specified event occurs, Cloud Eye automatically sends an alarm notification. This helps you quickly identify and address issues.

Precautions

As alarm rules and metric data grow quickly, the alarm engine faces higher computational demands, causing a continuous rise in alarm notifications. During sudden service changes or targeted attacks for large customers, the alarm link can get backlogged or fail entirely. Existing systems lack proper tools for identifying risks, protecting itself, and recovering fast, which delays fixing issues and worsens problems. To solve this, this system introduces overload protection features. It can track tenant-level alarms, spot unusual activities compared to past patterns, locate overload causes (for example, specific cloud services), limit or disable alarm notifications of those sources, and clear backlogged tasks. These features allow the system to identify overloads in 3 minutes, locate their sources in 5 minutes, and recover fully within 10 minutes, improving its ability to handle overloads and extreme situations effectively.

Cloud Eye limits or disables alarm rules if they are triggered often due to inappropriate settings. This happens when:

  • Rate limiting rules: Cloud Eye provides automatic rate limiting based on the combination of alarm rule IDs and alarm policy IDs to prevent frequent alarms caused by inappropriate alarm rule settings.
    • Rate limiting granularity: performed independently based on the combination of alarm rule IDs and alarm policy IDs. Rate limiting of different policies under the same rule does not affect each other. For example, if an alarm rule has three policies and one of them triggers alarms frequently due to a low threshold, only this policy will be affected by rate limiting. The other two policies will continue to generate alarms as usual.
    • Rate limiting policy: The system automatically sets the maximum number of alarms that can be generated per minute based on the processing capability of the current cluster. If the number of alarms triggered by an alarm policy exceeds the threshold, no additional alarm records will be generated, but the raw data of metrics will not be discarded. Alarms will be generated automatically in the next period.
  • Disabling mechanism: Rate limiting serves as the first layer of protection and automatically takes effect at the early stage of an alarm storm to prevent backlog and spread. If the issue persists (for example, continuous overload), O&M personnel will disable the alarm rules or policies of the overloaded sources to trigger the full-link fast discarding mechanism.
  • Typical scenarios where rate limiting or disabling is triggered:
    • There are too many resources monitored, and the alarm thresholds are improper.
    • Too many alarm rules target the same resources or metrics.