Alerting

Alerting mechanism measures system performance according to the metrics of services/instances/endpoints from different layers. Alerting kernel is an in-memory, time-window based queue.

The alerting core is driven by a collection of rules defined in config/alarm-settings.yml. There are three parts to alerting rule definitions.

  1. alerting rules. They define how metrics alerting should be triggered and what conditions should be considered.
  2. hooks. The list of hooks, which should be called after an alerting is triggered.

Entity name

Defines the relation between scope and entity name.

  • Service: Service name
  • Instance: {Instance name} of {Service name}
  • Endpoint: {Endpoint name} in {Service name}
  • Service Relation: {Source service name} to {Dest service name}
  • Instance Relation: {Source instance name} of {Source service name} to {Dest instance name} of {Dest service name}
  • Endpoint Relation: {Source endpoint name} in {Source Service name} to {Dest endpoint name} in {Dest service name}

Rules

An alerting rule is made up of the following elements:

  • Rule name. A unique name shown in the alarm message. It must end with _rule.
  • Expression. A MQE expression that defines the conditions of the rule. The result type must be SINGLE_VALUE and the root operation of the expression must be a Compare Operation which provides 1(true) or 0(false) result. When the result is 1(true), the alarm will be triggered. For example, avg(service_resp_time / 1000) > 1 is a valid expression to indicate the request latency is slower than 1s. The typical illegal expressions are
    • avg(service_resp_time > 1000) + 1 expression root doesn’t use Compare Operation
    • service_resp_time > 1000 expression return a TIME_SERIES_VALUES type of values rather than a SINGLE_VALUE value.

The metrics names in the expression could be found in the list of all potential metrics name doc.

  • Include names. Entity names that are included in this rule. Please follow the entity name definitions.
  • Exclude names. Entity names that are excluded from this rule. Please follow the entity name definitions.
  • Include names regex. A regex that includes entity names. If both include-name list and include-name regex are set, both rules will take effect.
  • Exclude names regex. A regex that excludes entity names. Both rules will take effect if both include-label list and include-label regex are set.
  • Tags. Tags are key/value pairs that are attached to alarms. Tags are used to specify distinguishing attributes of alarms that are meaningful and relevant to users. If you want to make these tags searchable on the SkyWalking UI, you may set the tag keys in core/default/searchableAlarmTags or through the system environment variable SW_SEARCHABLE_ALARM_TAG_KEYS. The key level is supported by default.
  • Period. The size of metrics cache in minutes for checking the alarm conditions. This is a time window that corresponds to the backend deployment env time.
  • Hooks. Binding the specific names of the hooks when the alarm is triggered. The name format is {hookType}.{hookName} (slack.custom1 e.g.) and must be defined in the hooks section of the alarm-settings.yml file. If the hook name is not specified, the global hook will be used.
  • Silence period. After the alarm is triggered at Time-N (TN), there will be silence during the TN -> TN + period. By default, it works in the same manner as period. The same Alarm (having the same ID in the same metrics name) may only be triggered once within a period.

Such as for a metric, there is a shifting window as following at T7.

T1 T2 T3 T4 T5 T6 T7
Value1 Value2 Value3 Value4 Value5 Value6 Value7
  • Period(Time point T1 ~ T7) are continuous data points for minutes. Notice, alerts are not supported above minute-by-minute periods as they would not be efficient.
  • Values(Value1 ~ Value7) are the values or labeled values for every time point.
  • Expression is calculated based on the metric values(Value1 ~ Value7). For example, expression avg(service_resp_time) > 1000, if the value are 1001, 1001, 1001, 1001, 1001, 1001, 1001, the calculation is ((1001 + 10001 + ... + 1001) / 7) > 1000 and the result would be 1(true). Then the alarm would be triggered.
  • In every minute, the window would shift automatically. At T8, Value8 would be cached, and T1/Value1 would be removed from the window.

NOTE:

  • If the expression include labeled metrics and result has multiple labeled value(e.g. sum(service_percentile{p='50,75'} > 1000) >= 3), the alarm will be triggered if any of the labeled value result matches 3 times of the condition(P50 > 1000 or P75 > 1000).
  • One alarm rule is targeting the same entity level, such as service-level expression (avg(service_resp_time) > 1000). Set entity names(Include/Exclude names…) according to metrics entity levels, do not include different entity levels metrics in the same expression, such as service metrics and endpoint metrics.
rules:
  # Rule unique name, must be ended with `_rule`.
  endpoint_percent_rule:
    # A MQE expression and the root operation of the expression must be a Compare Operation.
    expression: sum((endpoint_sla / 100) < 75) >= 3
    # The length of time to evaluate the metrics
    period: 10
    # How many times of checks, the alarm keeps silence after alarm triggered, default as same as period.
    silence-period: 10
    message: Successful rate of endpoint {name} is lower than 75%
    tags:
      level: WARNING
  service_percent_rule:
    expression: sum((service_sla / 100) < 85) >= 4
    # [Optional] Default, match all services in this metrics
    include-names:
      - service_a
      - service_b
    exclude-names:
      - service_c
    period: 10
    message: Service {name} successful rate is less than 85%
  service_resp_time_percentile_rule:
    expression: sum(service_percentile{p='50,75,90,95,99'} > 1000) >= 3
    period: 10
    silence-period: 5
    message: Percentile response time of service {name} alarm in 3 minutes of last 10 minutes, due to more than one condition of p50 > 1000, p75 > 1000, p90 > 1000, p95 > 1000, p99 > 1000
  meter_service_status_code_rule:
    expression: sum(aggregate_labels(meter_status_code{status='4xx,5xx'},sum) > 10) > 3
    period: 10
    silence-period: 5
    message: The request number of entity {name} 4xx and 5xx status is more than expected.
    hooks:
      - "slack.custom1"
      - "pagerduty.custom1"
  comp_rule:
    expression: (avg(service_sla / 100) > 80) * (avg(service_percentile{p='0'}) > 1000) == 1
    period: 10
    message: Service {name} avg successful rate is less than 80% and P50 of avg response time is over 1000ms in last 10 minutes.
    tags:
      level: CRITICAL
    hooks:
      - "slack.default"
      - "slack.custom1"
      - "pagerduty.custom1"

Default alarm rules

For convenience’s sake, we have provided a default alarm-setting.yml in our release. It includes the following rules:

  1. Service average response time over 1s in the last 3 minutes.
  2. Service success rate lower than 80% in the last 2 minutes.
  3. Percentile of service response time over 1s in the last 3 minutes
  4. Service Instance average response time over 1s in the last 2 minutes, and the instance name matches the regex.
  5. Endpoint average response time over 1s in the last 2 minutes.
  6. Database access average response time over 1s in the last 2 minutes.
  7. Endpoint relation average response time over 1s in the last 2 minutes.

List of all potential metrics name

The metrics names are defined in the official OAL scripts and MAL scripts.

Currently, metrics from the Service, Service Instance, Endpoint, Service Relation, Service Instance Relation, Endpoint Relation scopes could be used in Alarm, and the Database access scope is the same as Service.

Submit an issue or a pull request if you want to support any other scopes in Alarm.

Hooks

Hooks are a way to send alarm messages to the outside world. SkyWalking supports multiple hooks of the same type, each hook can support different configurations. For example, you can configure two Slack hooks, one named default and set is-default: true means this hook will apply on all Alarm Rules without config hooks. Another named custom1 will only apply on the Alarm Rules which with config hooks and include the name slack.custom1.

hooks:
  slack:
    # default here is just a name, set the field 'is-default: true' if this notification hook is expected to be default globally.
    default:
      # If true, this hook will apply on all rules, unless a rule has its own specific hook. Could have more than one default hooks in the same hook type.
      is-default: true 
      text-template: |-
        {
          "type": "section",
          "text": {
            "type": "mrkdwn",
            "text": ":alarm_clock: *Apache Skywalking Alarm* \n **%s**."
          }
        }        
      webhooks:
        - https://hooks.slack.com/services/x/y/zssss
    custom1:
      text-template: |-
        {
          "type": "section",
          "text": {
            "type": "mrkdwn",
            "text": ":alarm_clock: *Apache Skywalking Alarm* \n **%s**."
          }
        }        
      webhooks:
        - https://hooks.slack.com/services/x/y/custom1

Currently, SkyWalking supports the following hook types:

Webhook

The Webhook requires the peer to be a web container. The alarm message will be sent through HTTP post by application/json content type after you have set up Webhook hooks as follows:

webhook:
  default:
    is-default: true
    urls:
      - http://ip:port/xxx
      - http://ip:port/yyy

The JSON format is based on List<org.apache.skywalking.oap.server.core.alarm.AlarmMessage> with the following key information:

  • scopeId, scope. All scopes are defined in org.apache.skywalking.oap.server.core.source.DefaultScopeDefine.
  • name. Target scope entity name. Please follow the entity name definitions.
  • id0. The ID of the scope entity that matches with the name. When using the relation scope, it is the source entity ID.
  • id1. When using the relation scope, it is the destination entity ID. Otherwise, it is empty.
  • ruleName. The rule name configured in alarm-settings.yml.
  • alarmMessage. The alarm text message.
  • startTime. The alarm time measured in milliseconds, which occurs between the current time and the midnight of January 1, 1970 UTC.
  • tags. The tags configured in alarm-settings.yml.

See the following example:

[{
  "scopeId": 1, 
  "scope": "SERVICE",
  "name": "serviceA", 
  "id0": "12",  
  "id1": "",  
    "ruleName": "service_resp_time_rule",
  "alarmMessage": "alarmMessage xxxx",
  "startTime": 1560524171000,
    "tags": [{
        "key": "level",
        "value": "WARNING"
     }]
}, {
  "scopeId": 1,
  "scope": "SERVICE",
  "name": "serviceB",
  "id0": "23",
  "id1": "",
    "ruleName": "service_resp_time_rule",
  "alarmMessage": "alarmMessage yyy",
  "startTime": 1560524171000,
    "tags": [{
        "key": "level",
        "value": "CRITICAL"
    }]
}]

gRPC

The alarm message will be sent through remote gRPC method by Protobuf content type after you have set up gRPC hooks as follows:

gRPC:
  default:
    is-default: true
    target-host: ip
    target-port: port

The message contains key information which are defined in oap-server/server-alarm-plugin/src/main/proto/alarm-hook.proto.

Part of the protocol looks like this:

message AlarmMessage {
    int64 scopeId = 1;
    string scope = 2;
    string name = 3;
    string id0 = 4;
    string id1 = 5;
    string ruleName = 6;
    string alarmMessage = 7;
    int64 startTime = 8;
    AlarmTags tags = 9;
}

message AlarmTags {
    // String key, String value pair.
    repeated KeyStringValuePair data = 1;
}

message KeyStringValuePair {
    string key = 1;
    string value = 2;
}

Slack Chat

Follow the Getting Started with Incoming Webhooks guide and create new Webhooks.

The alarm message will be sent through HTTP post by application/json content type if you have configured Slack Incoming Webhooks as follows:

slack:
  default:
    is-default: true
    text-template: |-
      {
        "type": "section",
        "text": {
          "type": "mrkdwn",
          "text": ":alarm_clock: *Apache Skywalking Alarm* \n **%s**."
        }
      }      
    webhooks:
    - https://hooks.slack.com/services/x/y/z

WeChat

Note that only the WeChat Company Edition (WeCom) supports WebHooks. To use the WeChat WebHook, follow the Wechat Webhooks guide. The alarm message will be sent through HTTP post by application/json content type after you have set up Wechat Webhooks as follows:

wechat:
  default:
    is-default: true
    text-template: |-
      {
        "msgtype": "text",
        "text": {
          "content": "Apache SkyWalking Alarm: \n %s."
        }
      }      
    webhooks:
    - https://qyapi.weixin.qq.com/cgi-bin/webhook/send?key=dummy_key

DingTalk

Follow the Dingtalk Webhooks guide and create new Webhooks. You can configure an optional secret for an individual webhook URL for security purposes. The alarm message will be sent through HTTP post by application/json content type if you have configured DingTalk Webhooks as follows:

dingtalk:
  default:
    is-default: true
    text-template: |-
      {
        "msgtype": "text",
        "text": {
          "content": "Apache SkyWalking Alarm: \n %s."
        }
      }      
    webhooks:
    - url: https://oapi.dingtalk.com/robot/send?access_token=dummy_token
      secret: dummysecret

Feishu

Follow the Feishu Webhooks guide and create new Webhooks. You can configure an optional secret for an individual webhook URL for security purposes. If you want to direct a text to a user, you can configure ats, which is Feishu’s user_id and separated by “,” . The alarm message will be sent through HTTP post by application/json content type if you have configured Feishu Webhooks as follows:

feishu:
  default:
    is-default: true
    text-template: |-
      {
        "msg_type": "text",
        "content": {
          "text": "Apache SkyWalking Alarm: \n %s."
        },
        "ats":"feishu_user_id_1,feishu_user_id_2"
      }      
    webhooks:
    - url: https://open.feishu.cn/open-apis/bot/v2/hook/dummy_token
      secret: dummysecret

Follow the WeLink Webhooks guide and create new Webhooks. The alarm message will be sent through HTTP post by application/json content type if you have configured WeLink Webhooks as follows:

welink:
  default:
    is-default: true
    text-template: "Apache SkyWalking Alarm: \n %s."
    webhooks:
    # you may find your own client_id and client_secret in your app, below are dummy, need to change.
    - client-id: "dummy_client_id"
      client-secret: dummy_secret_key
      access-token-url: https://open.welink.huaweicloud.com/api/auth/v2/tickets
      message-url: https://open.welink.huaweicloud.com/api/welinkim/v1/im-service/chat/group-chat
      # if you send to multi group at a time, separate group_ids with commas, e.g. "123xx","456xx"
      group-ids: "dummy_group_id"
      # make a name you like for the robot, it will display in group
      robot-name: robot

PagerDuty

The PagerDuty hook is based on Events API v2.

Follow the Getting Started section to create an Events API v2 integration on your PagerDuty service and copy the integration key.

Then configure as follows:

pagerduty:
  default:
    is-default: true
    text-template: "Apache SkyWalking Alarm: \n %s."
    integration-keys:
    - 5c6d805c9dcf4e03d09dfa81e8789ba1

You can also configure multiple integration keys.

Discord

Follow the Discord Webhooks guide and create a new webhook.

Then configure as follows:

discord:
  default:
    is-default: true
    text-template: "Apache SkyWalking Alarm: \n %s."
    webhooks:
    - url: https://discordapp.com/api/webhooks/1008166889777414645/8e0Am4Zb-YGbBqqbiiq0jSHPTEEaHa4j1vIC-zSSm231T8ewGxgY0_XUYpY-k1nN4HBl
      username: robot

Update the settings dynamically

Since 6.5.0, the alerting settings can be updated dynamically at runtime by Dynamic Configuration, which will override the settings in alarm-settings.yml.

In order to determine whether an alerting rule is triggered or not, SkyWalking needs to cache the metrics of a time window for each alerting rule. If any attribute (expression, period, etc.) of a rule is changed, the sliding window will be destroyed and re-created, causing the Alarm of this specific rule to restart again.

Keys with data types of alerting rule configuration file

Alerting element Configuration property key Type Description
Expression expression string MQE expression
Include names include-names string array
Exclude names exclude-names string array
Include names regex include-names-regex string Java regex Pattern
Exclude names regex exclude-names-regex string Java regex Pattern
Tags tags key-value pair
Period Period int
Silence period silence-period int
Message message string
Hooks hooks string array