Skip to content

Latest commit

 

History

History
392 lines (357 loc) · 12.8 KB

governance.md

File metadata and controls

392 lines (357 loc) · 12.8 KB
Raw

Service Governance Model

1. Unified Governance Hierarchy

Unified governance strategy configuration hierarchy for HTTP and traditional RPC.

.
└── Service
    ├── Group*
    │   ├── Path*
    │   │   ├── Method*

Service governance strategies are placed on groups, paths, and methods. They can be set step by step, with lower levels inheriting the configurations of higher levels by default.

The default strategy for a service is set on the default group default.

Type Service Group Path Method
HTTP Domain Group URL Path HTTP Method
RPC App-Level Registration App Name Group Interface Name Method Name
RPC Interface-Level Registration Interface Name Group / Method Name

2. Service Strategies

2.1 Multi-Active Service Strategy

Element Description
Write Protect Whether to write methods, needs to be disabled during disaster recovery
Variable Expression In unitization scenarios, can set routing variables from method parameters
Unit Policy Non-unitized, Local Unit Preferred, Unitized and Central
Default Unit Fault Tolerance Threshold Under the local unit preferred strategy, cross-unit access is allowed when unit metrics are below this threshold
Unit Fault Tolerance Threshold Set the fault tolerance threshold for each unit
Partition Policy Local Partition Preferred, Unlimited
Default Partition Fault Tolerance Threshold Under the local partition preferred strategy, cross-unit access is allowed when partition metrics are below this threshold
Partition Fault Tolerance Threshold Set the fault tolerance threshold for each partition 
{
  "writeProtect": false,
  "variableExpression": null,
  "unitPolicy": "PREFER_LOCAL_UNIT",
  "defaultUnitThreshold": 1,
  "unitRemotes": [
    {
      "name": "unit2",
      "type": "INSTANCES",
      "threshold": 1
    }
  ],
  "cellPolicy": "PREFER_LOCAL_CELL",
  "defaultCellThreshold": 3,
  "cellRemotes": [
    {
      "name": "cell1",
      "type": "INSTANCES",
      "threshold": 4
    }
  ]
}

2.2 Service Lane Strategy

Element Description
Auto Join Lane Whether to automatically join the lane routing when the application is deployed to a certain lane.
true Automatically join;
false Do not automatically join, follow the configured lane strategy;
null Inherit the configuration from the parent or lane, with the default lane configuration being to automatically join
Lane Space Configure participation in specific lane spaces and lane routes 
{
  "autoLaneEnabled": true,
  "lanePolicies": [
    {
      "laneSpaceId": "lane-space-1",
      "lanes": {
        "gray": "beta"
      }
    }
  ]
}

This strategy describes automatically joining lanes. In the lane space lane-space-1, it redirects the traffic of lane gray to lane beta.

2.3 Service Cluster Strategy

Element Description
Type Failover, Fast Fail, Fail Tolerance
Retry Policy When type is failover, configure retry policy including retry count, retry interval, timeout, retry statuses, and retry exceptions 
{
  "clusterPolicy": {
    "type": "failover",
    "retryPolicy": {
      "retry": 10,
      "retryInterval": 1000,
      "timeout": 5000,
      "codePolicy": {
        "parser": "JsonPath",
        "expression": "$.code",
        "contentTypes": [
          "application/json"
        ]
      },
      "retryStatuses": [
        500,
        502
      ],
      "retryExceptions": [
        "java.lang.NullPointerException"
      ]
    }
  }
}

2.4 Service Load Balancing Strategy

Element Description
Type Random weight and round-robin
1. RANDOM: Random weight;
2. ROUND_ROBIN: Round-robin
Sticky Sticky strategy
1. NONE: Not enabled;
2. PREFERRED: Prefer to keep sticky nodes;
3. FIXED: Fixed sticky nodes		 
{
  "loadBalancePolicy": {
    "policyType": "ROUND_ROBIN",
    "stickyType": "NONE"
  }
}

2.5 Service Rate Limiting Strategy

Element Description
Name Rate limiting strategy name
Type Rate limiting implementation type, including Resilience4j and TokenBucket
Sliding Windows Configure sliding window information including call count and time window
Max Wait Time Maximum wait time
Parameters Parameters required for rate limiting operations
Conditions Conditions for enabling rate limiting 
{
  "rateLimitPolicies": [
    {
      "name": "limit-rule-1",
      "version": 1704038400000,
      "realizeType": "Resilience4j",
      "slidingWindows": [
        {
          "threshold": 1,
          "timeWindowInMs": 1000
        },
        {
          "threshold": 3,
          "timeWindowInMs": 2000
        }
      ],
      "maxWaitMs": 100,
      "actionParameters": {
      },
      "relationType": "AND",
      "conditions": [
        {
          "type": "header",
          "opType": "EQUAL",
          "key": "x-live-ext-demo",
          "value": [
            "abc"
          ]
        }
      ]
    }
  ]
}

2.6 Service Concurrency Control Strategy

Element Description
Name Concurrency control strategy name
Type Rate limiting implementation type, including Resilience4j
Max Concurrency Maximum concurrency
Max Wait Time Maximum wait time
Parameters Parameters required for rate limiting operations
Conditions Conditions for enabling rate limiting 
{
  "concurrencyLimitPolicies": [
    {
      "name": "limit-rule-2",
      "version": 1704038400000,
      "realizeType": "Resilience4j",
      "maxConcurrency": 10,
      "maxWaitMs": 100,
      "actionParameters": {
      },
      "relationType": "AND",
      "conditions": [
        {
          "type": "header",
          "opType": "EQUAL",
          "key": "x-live-ext-demo",
          "value": [
            "abc"
          ]
        }
      ]
    }
  ]
}

2.7 Service Tag Routing Strategy

Element Description
Name Strategy name
Rules Define multiple rules, each including enabling conditions, target weights, and target tags 
{
  "routePolicies": [
    {
      "name": "route1",
      "tagRules": [
        {
          "order": 5,
          "relationType": "AND",
          "conditions": [
            {
              "opType": "EQUAL",
              "type": "header",
              "key": "x-live-unit",
              "values": [
                "unit1"
              ]
            }
          ],
          "destinations": [
            {
              "weight": 100,
              "relationType": "AND",
              "conditions": [
                {
                  "opType": "EQUAL",
                  "key": "unit",
                  "values": [
                    "unit1"
                  ]
                }
              ]
            }
          ]
        }
      ]
    }
  ]
}

2.8 Service Circuit Breaker Strategy

stateDiagram-v2
    Closed --> Open : fail [threshold reached]
    Closed --> Closed : fail [under threshold]
    Open --> HalfOpen : reset timeout
    HalfOpen --> Closed : success
    HalfOpen --> Open : fail
    
    Closed --> Closed : success
    Open --> Open : call / raise circuit open
Loading
Element Description
name Policy name
level SERVICE: service level; API: API level; INSTANCE: instance level
slidingWindowType Specify the type of sliding window, count: number of times; time: duration
slidingWindowSize Specify the size of the sliding window, if it is count, it represents the number of calls; if it is time, it represents seconds
minCallsThreshold Protect the threshold to prevent applications with too few calls from experiencing a fuse due to occasional failures
codePolicy Used to extract error codes from response body
errorCodes If one of the response codes in the list is returned, it will be recorded as a failure by the circuit breaker
failureRateThreshold Failure rate threshold to trigger circuit breaker
slowCallRateThreshold Threshold for the slow call count ratio that triggers circuit breaking
slowCallDurationThreshold How long does a call take to be considered slow
waitDurationInOpenState When the circuit breaker is triggered (status is open), how long should access tokens not be granted
allowedCallsInHalfOpenState When the circuit breaker enters the half-open state, how many attempts can be allowed for trial access
forceOpen Forced to open the fuse
degradeConfig When a fuse occurs, if the degradation configuration is performed, the configuration data will be returned as a response 
{
  "circuitBreakPolicies": [
    {
      "name": "cb1",
      "level": "SERVICE",
      "slidingWindowType": "count",
      "slidingWindowSize": 5,
      "minCallsThreshold": 1,
      "codePolicy": {
        "parser": "JsonPath",
        "expression": "$.code",
        "contentTypes": [
          "application/json"
        ]
      },
      "errorCodes": [
        "500",
        "502"
      ],
      "failureRateThreshold": 20,
      "slowCallRateThreshold": 20,
      "slowCallDurationThreshold": 1000,
      "waitDurationInOpenState": 50,
      "allowedCallsInHalfOpenState": 3,
      "forceOpen": false,
      "degradeConfig": {
        "responseCode": 200,
        "attributes": {
          "degrade-header-1": "degraded1",
          "degrade-header-2": "degraded2"
        },
        "responseBody": "Hello, your request has been downgraded."
      },
      "version": 1704038400000
    }
  ]
}

3. Complete Service Strategy Skeleton

Below is the complete service strategy in JSON format.

[
  {
    "name": "service-provider",
    "serviceType": "HTTP",
    "version": 0,
    "groups": [
      {
        "name": "default",
        "defaultGroup": true,
        "servicePolicy": {
          "livePolicy": {
          },
          "lanePolicies": [
          ],
          "clusterPolicy": {
          },
          "loadBalancePolicy": {
          },
          "rateLimitPolicies": [
          ],
          "concurrencyLimitPolicies": [
          ],
          "routePolicies": [
          ],
          "circuitBreakPolicies": [
          ]
        },
        "paths": [
          {
            "path": "/echo",
            "matchType": "EQUAL",
            "servicePolicy": {
            },
            "methods": [
              {
                "name": "GET",
                "servicePolicy": {
                }
              }
            ]
          }
        ]
      }
    ]
  }
]