Add Sentinel MetricExtension, which provides extension to Sentinel internal statistics.
This extension provides callbacks when a request passes rule checking,
blocked by flow control, successfully end or exception occurred.
Accompanied by these events, response time and current thread number
will be recorded too.
Signed-off-by: Carpenter Lee <hooleeucas@163.com>
- Refactor the semantics and logic for blockHandler and fallback
- Add `fallbackClass` support for fallback in global classes
- Add `defaultFallback` support
- Update test cases and demo
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
- The `appType` can be retrieved from `csp.sentinel.app.type` field. It's useful to identify the service type (e.g. API gateway).
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
* Rename: MetricsLeapArray -> BucketLeapArray
* Add implementation for `FutureBucketLeapArray`, a kind of `BucketLeapArray` that only reserves for future buckets, which is used for calculating occupied future tokens.
* Add OccupiableBucketLeapArray that combines common BucketLeapArray with FutureBucketLeapArray. The rollingNumberInSecond in StatisticNode now uses OccupiableBucketLeapArray by default.
* Add OccupySupport interface. Node now implements OccupySupport interface.
* Add occupy-related methods in Metric and ArrayMetric.
* Handle prioritized requests in default traffic shaping controller.
* Update default occupyTimeout to 500ms
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
- Update test cases (use assertEquals(e, a, delta) instead)
- Also add `totalPass` method in Node interface
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
- `Entry` class now implements `AutoCloseable` so it supports try-with-resources (`close` method adapts to `exit`)
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
- Refactor token client common config and assign config
- Log enhancement when transport to token server failed
- Add `getState` method to `ClusterTokenClient` interface
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
- Support stop cluster mode (`NOT-STARTED` mode)
- Fix bug when updating cluster state via command API (should modify via state property)
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
- Add `UnaryLeapArray` and `RequestLimiter` to enable simple QPS limit
- Improve cluster rule manager and server config manager to support request limiter
- Support `TOO_MANY_REQUEST` status in client side
- Also improve the automatic namespace register of embedded server mode
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
- Refactor MetricBucket to support add multiple count
- Refactor Node and Metric related classes
- Refactor for StatisticSlot
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
* Fix#330 : fix listMetricFiles() returns wrong file name list.
Signed-off-by: Carpenter Lee <hooleeucas@163.com>
* Add timestamp check when read metrics.
Signed-off-by: Carpenter Lee <hooleeucas@163.com>
* Refine test case
Signed-off-by: Carpenter Lee <hooleeucas@163.com>
- Code refinement for heartbeat component in transport related module
- Add remove support of `SentinelConfig`
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
- The constructor now accept `sampleCount` and `windowIntervalMs` so that it can match the two basic properties
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
- Remove "borrow-from-ref" mode
- Improve flow checker to support both embedded server mode and client mode
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
- Add a `ClusterTransportClient` for transport abstraction of Sentinel cluster
- Add universal request/response interface and common ClusterRequest/ClusterResponse abstraction
- Add common request/response data entity
- Add basic abstraction of codec (EntityWriter and EntityDecoder)
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
- Update ParamFlowRule to support cluster mode
- Add `ParamFlowClusterConfig` to provide cluster mode items for the rule
- Update ParamFlowChecker to support cluster flow mode
- Extract ParamFlowRuleUtil class
- Change type of `flowId` from Integer to Long
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
- Add new field `clusterMode` and `clusterConfig` for cluster mode
- Add a `ClusterFlowConfig` class for specific items for cluster flow control
- Update FlowRuleChecker to support cluster mode
- Extract valid rule checking and rule map generating logic to FlowRuleUtil
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
- Add a universal `TokenService` SPI interface for both local flow control and distributed flow control
- Add TokenResult entity to represents result of acquiring token
- Add `ClusterTokenClient` as the SPI interface for client of Sentinel cluster flow control
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
- Update ParamFlowRule to support cluster mode
- Add `ParamFlowClusterConfig` to provide cluster mode items for the rule
- Update ParamFlowChecker to support cluster flow mode
- Extract ParamFlowRuleUtil class
- Change type of `flowId` from Integer to Long
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
- Add new field `clusterMode` and `clusterConfig` for cluster mode
- Add a `ClusterFlowConfig` class for specific items for cluster flow control
- Update FlowRuleChecker to support cluster mode
- Extract valid rule checking and rule map generating logic to FlowRuleUtil
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
- Add a universal `TokenService` SPI interface for both local flow control and distributed flow control
- Add TokenResult entity to represents result of acquiring token
- Add `ClusterTokenClient` as the SPI interface for client of Sentinel cluster flow control
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
- Enhance rule checking in rule managers and Sentinel Dashboard frontend
- Enhance error information when reporting invalid fields in rule dialog
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
* sentinel-transport-netty-http will also retry when port specified has been used
* LogBase support variable arguments
* Remove duplicated JVM properties overriding in SentinelConfig
* Add VersionUtil to get version from jar
- Add callback registry for statistic slot for extensions.
- Add a new module `sentinel-parameter-flow-control` under `sentinel-extension`.
- Add a `CacheMap` interface to provide abstraction for cache. We use ConcurrentLinkedHashMap as the default implementation (LRU strategy)..
- Add a `ParameterMetric` class as frequent parameter metrics for a specific resource. The metric map is located in `ParamFlowSlot` rather than `ClusterNode`.
- Implement `ParameterLeapArray` as statistic data structure for frequent parameters in a period of time window.
- Add `ParamFlowSlot` as the checker slot; Add `ParamFlowChecker` to do rule checking; Add `ParamFlowRuleManager` to do rule managing.
- The statistic metrics for frequent parameters is enabled only if the related resource has configured parameter flow rule; Parameter metrics for removed rules will be cleared automatically.
- Leverage extensible `SlotChainBuilder` to provide a `HotParamSlotChainBuilder`.
- Add command handlers for hot param rules.
- Add test cases and demo.
- When amount of context exceeds the threshold, the `NullContext` will be set to current thread local. Thus, when checking context in `CtSph#entry`, once `NullContext` detected, the entry will not do rule checking on the slot chain.
- Cache the default context during initialization. Then when amount of context exceeds the threshold, entries under default context can do rule checking under default context.
- Enhance the error message
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
- Add several `asyncEntry` method in `Sph` interface and `SphU` class.
- Add a new `AsyncEntry` class for asynchronous entry representation. Some refactor for `CtEntry`.
- Refactored `CtSph` and implement `asyncEntryInternal` for asynchronous entry.
- Add `runOnContext` and `replaceContext` method in `ContextUtil` for context switching.
- Support extensible `SlotChainBuilder` using SPI mechanism
- Add a `SlotChainProvider` to load slot chain builder and create new slot chains
Signed-off-by: Eric Zhao <sczyh16@gmail.com>
Lin.Liang