Skip to content

Latest commit

 

History

History
460 lines (425 loc) · 16.8 KB

API.md

File metadata and controls

460 lines (425 loc) · 16.8 KB
Raw

Autotune REST API

The Autotune REST API design is proposed as follows:

listStacks

Get the list of application stacks monitored by autotune.

Request GET /listStacks

GET /listStacks?experiment_name=<EXPERIMENT_NAME>

curl -H 'Accept: application/json' http://<URL>:<PORT>/listStacks?experiment_name=<EXPERIMENT_NAME>

Response

[
    {
        "experiment_name": "galaxies-autotune-min-http-response-time",
        "experiment_id": "7c07cf4db16adcf76bad79394c9e7df2f3b8d8e6942cfa3f7b254b5aec1299b0",
        "objective_function": "request_sum/request_count",
        "hpo_algo_impl": "optuna_tpe",
        "stacks": ["dinogun/galaxies:1.2-jdk-11.0.10_9"],
        "slo_class": "response_time",
        "direction": "minimize"
    },
    {
        "experiment_name": "petclinic-autotune-max-http-throughput",
        "experiment_id": "629ad1c575dd81576c98142aa6de9ddc241de0a9f008586923f200b3a6bc83c6",
        "objective_function": "request_count",
        "hpo_algo_impl": "optuna_tpe",
        "stacks": ["kruize/spring_petclinic:2.2.0-jdk-11.0.8-openj9-0.21.0"],
        "slo_class": "throughput",
        "direction": "maximize"
    }
]

listStackLayers

Returns the list of application stacks monitored by autotune along with layers detected in the stacks.

Request GET /listStackLayers

GET /listStackLayers?experiment_name=<EXPERIMENT_NAME>

curl -H 'Accept: application/json' http://<URL>:<PORT>/listStackLayers?experiment_name=<EXPERIMENT_NAME>

Response

[
    {
        "experiment_name": "galaxies-autotune-min-http-response-time",
        "experiment_id": "7c07cf4db16adcf76bad79394c9e7df2f3b8d8e6942cfa3f7b254b5aec1299b0",
        "objective_function": "request_sum/request_count",
        "hpo_algo_impl": "optuna_tpe",
        "stacks": [{
            "layers": [
                {
                    "layer_id": "af07fd998199bf2d57f95dc18f2cc2311b72f6de11e7e949b566fcdc5ecb443b",
                    "layer_level": 0,
                    "layer_name": "container",
                    "layer_details": "generic container tunables"
                },
                {
                    "layer_id": "63f4bd430913abffaa6c41a5e05015d5fea23134c99826470c904a7cfe56b40c",
                    "layer_level": 1,
                    "layer_name": "hotspot",
                    "layer_details": "hotspot tunables"
                },
                {
                    "layer_id": "3ec648860dd10049b2488f19ca6d80fc5b50acccdf4aafaedc2316c6eea66741",
                    "layer_level": 2,
                    "layer_name": "quarkus",
                    "layer_details": "quarkus tunables"
                }
            ],
            "stack_name": "dinogun/galaxies:1.2-jdk-11.0.10_9"
        }],
        "slo_class": "response_time",
        "direction": "minimize"
    },
    {
        "experiment_name": "petclinic-autotune-max-http-throughput",
        "experiment_id": "629ad1c575dd81576c98142aa6de9ddc241de0a9f008586923f200b3a6bc83c6",
        "objective_function": "request_count",
        "hpo_algo_impl": "optuna_tpe",
        "stacks": [{
            "layers": [
                {
                    "layer_id": "af07fd998199bf2d57f95dc18f2cc2311b72f6de11e7e949b566fcdc5ecb443b",
                    "layer_level": 0,
                    "layer_name": "container",
                    "layer_details": "generic container tunables"
                }
            ],
            "stack_name": "kruize/spring_petclinic:2.2.0-jdk-11.0.8-openj9-0.21.0"
        }],
        "slo_class": "throughput",
        "direction": "maximize"
    }
]

listStackTunables

Returns the list of application stacks monitored by autotune along with their tunables. Request GET /listStackTunables gives the tunables and layer information for all the application stacks monitored by autotune.

GET /listStackTunables?experiment_name=<EXPERIMENT_NAME> for getting the tunables information of a specific application.

GET /listStackTunables?experiment_name=<EXPERIMENT_NAME>&layer_name=<LAYER> for getting tunables of a specific layer for the application.

curl -H 'Accept: application/json' http://<URL>:<PORT>/listStackTunables?experiment_name=<EXPERIMENT_NAME>

Response

[
    {
        "experiment_name": "petclinic-autotune-max-http-throughput",
        "experiment_id": "629ad1c575dd81576c98142aa6de9ddc241de0a9f008586923f200b3a6bc83c6",
        "objective_function": "request_count",
        "hpo_algo_impl": "optuna_tpe",
        "stacks": [{
            "layers": [
                {
                    "layer_id": "af07fd998199bf2d57f95dc18f2cc2311b72f6de11e7e949b566fcdc5ecb443b",
                    "layer_level": 0,
                    "tunables": [
                        {
                            "value_type": "double",
                            "lower_bound": "150.0Mi",
                            "name": "memoryRequest",
                            "step": 1,
                            "query_url": "http://10.111.106.208:9090/api/v1/query?query=container_memory_working_set_bytes{container=\"\", pod=\"kruize/spring_petclinic:2.2.0-jdk-11.0.8-openj9-0.21.0\"}",
                            "upper_bound": "300.0Mi"
                        },
                        {
                            "value_type": "double",
                            "lower_bound": "1.0",
                            "name": "cpuRequest",
                            "step": 0.01,
                            "query_url": "http://10.111.106.208:9090/api/v1/query?query=(container_cpu_usage_seconds_total{container!=\"POD\", pod=\"kruize/spring_petclinic:2.2.0-jdk-11.0.8-openj9-0.21.0\"}[1m])",
                            "upper_bound": "3.0"
                        }
                    ],
                    "layer_name": "container",
                    "layer_details": "generic container tunables"
                }
            ],
            "stack_name": "kruize/spring_petclinic:2.2.0-jdk-11.0.8-openj9-0.21.0"
        }],
        "function_variables": [{
            "value_type": "double",
            "name": "request_count",
            "query_url": "http://10.111.106.208:9090/api/v1/query?query=rate(http_server_requests_seconds_count{exception=\"None\",method=\"GET\",outcome=\"SUCCESS\",status=\"200\",uri=\"/webjars/**\",}[1m])"
        }],
        "slo_class": "throughput",
        "direction": "maximize"
    }
]

listAutotuneTunables

Get the tunables supported by autotune for the SLO.

Request GET /listAutotuneTunables gives all tunables for all layers in the cluster

GET /listAutotuneTunables?slo_class=<SLO_CLASS> gives all tunables for the SLO class

GET /listAutotuneTunables?slo_class=<SLO_CLASS>&layer_name=<LAYER> gives tunables for the SLO class and the layer

curl -H 'Accept: application/json' http://<URL>:<PORT>/listAutotuneTunables?slo_class=<SLO_CLASS>

Response

[
    {
        "layer_id": "af07fd998199bf2d57f95dc18f2cc2311b72f6de11e7e949b566fcdc5ecb443b",
        "layer_level": 0,
        "tunables": [
            {
                "value_type": "double",
                "lower_bound": "150.0Mi",
                "name": "memoryRequest",
                "step": 1,
                "query_url": "http://10.111.106.208:9090/api/v1/query?query=container_memory_working_set_bytes{container=\"\", pod=\"$POD$\"}",
                "upper_bound": "300.0Mi"
            },
            {
                "value_type": "double",
                "lower_bound": "1.0",
                "name": "cpuRequest",
                "step": 0.01,
                "query_url": "http://10.111.106.208:9090/api/v1/query?query=(container_cpu_usage_seconds_total{container!=\"POD\", pod=\"$POD$\"}[1m])",
                "upper_bound": "3.0"
            }
        ],
        "layer_name": "container",
        "layer_details": "generic container tunables"
    },
    {
        "layer_id": "63f4bd430913abffaa6c41a5e05015d5fea23134c99826470c904a7cfe56b40c",
        "layer_level": 1,
        "tunables": [{
            "value_type": "integer",
            "lower_bound": "9",
            "name": "MaxInlineLevel",
            "step": 1,
            "query_url": "http://10.111.106.208:9090/api/v1/query?query=jvm_memory_used_bytes{area=\"heap\", container=\"\", pod=\"$POD$\"}",
            "upper_bound": "50"
        }],
        "layer_name": "hotspot",
        "layer_details": "hotspot tunables"
    },
    {
        "layer_id": "3ec648860dd10049b2488f19ca6d80fc5b50acccdf4aafaedc2316c6eea66741",
        "layer_level": 2,
        "tunables": [
            {
                "value_type": "integer",
                "lower_bound": "1",
                "name": "quarkus.thread-pool.core-threads",
                "step": 1,
                "query_url": "none",
                "upper_bound": "10"
            },
            {
                "value_type": "integer",
                "lower_bound": "1",
                "name": "quarkus.thread-pool.queue-size",
                "step": 1,
                "query_url": "none",
                "upper_bound": "100"
            },
            {
                "value_type": "integer",
                "lower_bound": "1",
                "name": "quarkus.hibernate-orm.jdbc.statement-fetch-size",
                "step": 1,
                "query_url": "none",
                "upper_bound": "50"
            }
        ],
        "layer_name": "quarkus",
        "layer_details": "quarkus tunables"
    }
]

SearchSpace

Generates the search space used for the analysis.

Request GET /searchSpace gives the search space for all application stacks monitored.

GET /searchSpace?experiment_name=<EXPERIMENT_NAME> gives the search space for a specific application stack.

curl -H 'Accept: application/json' http://<URL>:<PORT>/searchSpace

Response

[
    {
        "experiment_name": "galaxies-autotune-min-http-response-time",
        "experiment_id": "7c07cf4db16adcf76bad79394c9e7df2f3b8d8e6942cfa3f7b254b5aec1299b0",
        "objective_function": "request_sum/request_count",
        "hpo_algo_impl": "optuna_tpe",
        "tunables": [
            {
                "value_type": "double",
                "lower_bound": "150.0Mi",
                "name": "memoryRequest",
                "step": 1,
                "upper_bound": "300.0Mi"
            },
            {
                "value_type": "double",
                "lower_bound": "1.0",
                "name": "cpuRequest",
                "step": 0.01,
                "upper_bound": "3.0"
            },
            {
                "value_type": "integer",
                "lower_bound": "9",
                "name": "MaxInlineLevel",
                "step": 1,
                "upper_bound": "50"
            },
            {
                "value_type": "integer",
                "lower_bound": "1",
                "name": "quarkus.thread-pool.core-threads",
                "step": 1,
                "upper_bound": "10"
            },
            {
                "value_type": "integer",
                "lower_bound": "1",
                "name": "quarkus.thread-pool.queue-size",
                "step": 1,
                "upper_bound": "100"
            },
            {
                "value_type": "integer",
                "lower_bound": "1",
                "name": "quarkus.hibernate-orm.jdbc.statement-fetch-size",
                "step": 1,
                "upper_bound": "50"
            }
        ],
        "direction": "minimize"
    }
]

Health

Get the status of autotune.

Request GET /health

curl -H 'Accept: application/json' http://<URL>:<PORT>/health

Response

Healthy

CreateExperimentTrial

Create experiment trials using input JSON provided by Analyser module.

Request POST /createExperimentTrial

curl -H 'Accept: application/json' -X POST --data 'copy paste below JSON' http://<URL>:<PORT>/createExperimentTrial

{
    "settings": {
        "trial_settings": {
            "measurement_cycles": "3",
            "warmup_duration": "1min",
            "warmup_cycles": "3",
            "measurement_duration": "1min",
            "iterations": "3"
        },
        "deployment_settings": {
            "deployment_tracking": {
                "trackers": [
                    "training"
                ]
            },
            "deployment_policy": {
                "type": "rollingUpdate"
            }
        }
    },
    "experiment_name": "quarkus-resteasy-autotune-min-http-response-time-db",
    "deployments": {
        "training": {
            "pod_metrics": {
                "request_sum": {
                    "datasource": "prometheus",
                    "query": "rate(http_server_requests_seconds_sum{method=\"GET\",outcome=\"SUCCESS\",status=\"200\",uri=\"/db\",}[1m])",
                    "name": "request_sum"
                },
                "request_count": {
                    "datasource": "prometheus",
                    "query": "rate(http_server_requests_seconds_count{method=\"GET\",outcome=\"SUCCESS\",status=\"200\",uri=\"/db\",}[1m])",
                    "name": "request_count"
                }
            },
            "deployment_name": "tfb-qrh-sample",
            "namespace": "default",
            "containers": {
                "kruize/tfb-qrh:1.13.2.F_mm.v1": {
                    "image_name": "kruize/tfb-qrh:1.13.2.F_mm.v1",
                    "container_name": "tfb-server",
                    "container_metrics": {
                        "MaxInlineLevel": {
                            "datasource": "prometheus",
                            "query": "jvm_memory_used_bytes{area=\"heap\", $CONTAINER_LABEL$=\"\", $POD_LABEL$=\"$POD$\"}",
                            "name": "MaxInlineLevel"
                        },
                        "memoryRequest": {
                            "datasource": "prometheus",
                            "query": "container_memory_working_set_bytes{$CONTAINER_LABEL$=\"\", $POD_LABEL$=\"$POD$\"}",
                            "name": "memoryRequest"
                        },
                        "cpuRequest": {
                            "datasource": "prometheus",
                            "query": "(container_cpu_usage_seconds_total{$CONTAINER_LABEL$!=\"POD\", $POD_LABEL$=\"$POD$\"}[1m])",
                            "name": "cpuRequest"
                        }
                    },
                    "config": {
                        "0": {
                            "requests": {
                                "cpu": {
                                    "amount": "2.11",
                                    "format": "",
                                    "additionalProperties": {}
                                },
                                "memory": {
                                    "amount": "160.0",
                                    "format": "",
                                    "additionalProperties": {}
                                }
                            },
                            "env": [
                                {
                                    "name": "JAVA_OPTIONS",
                                    "additionalProperties": {},
                                    "value": " -server -XX:MaxRAMPercentage=70 -XX:+AllowParallelDefineClass -XX:MaxInlineLevel=21 -XX:+UseZGC -XX:+TieredCompilation -Dquarkus.thread-pool.queue-size=27 -Dquarkus.thread-pool.core-threads=9"
                                },
                                {
                                    "name": "JDK_JAVA_OPTIONS",
                                    "additionalProperties": {},
                                    "value": " -server -XX:MaxRAMPercentage=70 -XX:+AllowParallelDefineClass -XX:MaxInlineLevel=21 -XX:+UseZGC -XX:+TieredCompilation -Dquarkus.thread-pool.queue-size=27 -Dquarkus.thread-pool.core-threads=9"
                                }
                            ],
                            "limits": {
                                "cpu": {
                                    "amount": "2.11",
                                    "format": "",
                                    "additionalProperties": {}
                                },
                                "memory": {
                                    "amount": "160.0",
                                    "format": "",
                                    "additionalProperties": {}
                                }
                            }
                        }
                    }
                }
            },
            "type": "training"
        }
    },
    "experiment_id": "04c99daec35563782c29f17ebe568ea96065a7b20b93eb47c225a2f2ad769445",
    "datasource_info": {
        "name": "prometheus",
        "url": "http://10.101.144.137:9090"
    },
    "trial_info": {
        "trial_id": "",
        "trial_num": 0,
        "trial_result_url": "http://localhost:8080/listExperiments?experiment_name=quarkus-resteasy-autotune-min-http-response-time-db"
    },
    "namespace": "default"
}

Response

201