skywalking监控k8s skywalking监控指标

转载

JAVA小侠影 2024-02-10 13:03:01

文章标签 skywalking监控k8s 响应时间服务端 Endpoint 文章分类 运维

网上看了很多，发现对于Skywalking支持哪些指标名称metrics，官方文档跟博客几乎都是指明了一个路径，没有人详细的解释，支持哪些指标，这些指标的作用又有什么作用，导致大家自定义指标的时候有很多困难。

所以这里给大家总结下，如有错误，及时指正：

Skywalking的oap指标存放在：/apache-skywalking-apm-bin-es78/config/oal/*.oap 目录下

先来看第一个oap文件:

core.oal

1 / All scope metrics
 2 all_percentile = from(All.latency).percentile(10);  // Multiple values including p50, p75, p90, p95, p99
 3 all_heatmap = from(All.latency).histogram(100, 20); // 
 4 
 5 // Service scope metrics 服务
 6 service_resp_time = from(Service.latency).longAvg(); // 服务的平均响应时间
 7 service_sla = from(Service.*).percent(status == true); // 服务的请求成功率
 8 service_cpm = from(Service.*).cpm(); //服务的每分钟调用次数
 9 service_percentile = from(Service.latency).percentile(10); // Multiple values including p50, p75, p90, p95, p99
10 service_apdex = from(Service.latency).apdex(name, status); // 服务的应用性能指标，apdex的衡量的是衡量满意的响应时间与不满意的响应时间的比率，默认的请求满意时间是500ms
11 
12 // Service relation scope metrics for topology 服务与服务间调用的调用度量指标
13 service_relation_client_cpm = from(ServiceRelation.*).filter(detectPoint == DetectPoint.CLIENT).cpm();//在客户端检测到的每分钟调用次数
14 service_relation_server_cpm = from(ServiceRelation.*).filter(detectPoint == DetectPoint.SERVER).cpm();//在服务端检测到的每分钟调用的次数
15 service_relation_client_call_sla = from(ServiceRelation.*).filter(detectPoint == DetectPoint.CLIENT).percent(status == true);//在客户端检测到成功率
16 service_relation_server_call_sla = from(ServiceRelation.*).filter(detectPoint == DetectPoint.SERVER).percent(status == true);//在服务端检测到的成功率
17 service_relation_client_resp_time = from(ServiceRelation.latency).filter(detectPoint == DetectPoint.CLIENT).longAvg();//在客户端检测到的平均响应时间
18 service_relation_server_resp_time = from(ServiceRelation.latency).filter(detectPoint == DetectPoint.SERVER).longAvg();//在服务端检测到的平均响应时间
19 service_relation_client_percentile = from(ServiceRelation.latency).filter(detectPoint == DetectPoint.CLIENT).percentile(10); // Multiple values including p50, p75, p90, p95, p99
20 service_relation_server_percentile = from(ServiceRelation.latency).filter(detectPoint == DetectPoint.SERVER).percentile(10); // Multiple values including p50, p75, p90, p95, p99
21 
22 // Service Instance relation scope metrics for topology 服务实例与服务实例之间的调用度量指标
23 service_instance_relation_client_cpm = from(ServiceInstanceRelation.*).filter(detectPoint == DetectPoint.CLIENT).cpm();//在客户端实例检测到的每分钟调用次数
24 service_instance_relation_server_cpm = from(ServiceInstanceRelation.*).filter(detectPoint == DetectPoint.SERVER).cpm();//在服务端实例检测到的每分钟调用次数
25 service_instance_relation_client_call_sla = from(ServiceInstanceRelation.*).filter(detectPoint == DetectPoint.CLIENT).percent(status == true);//在客户端实例检测到的成功率
26 service_instance_relation_server_call_sla = from(ServiceInstanceRelation.*).filter(detectPoint == DetectPoint.SERVER).percent(status == true);//在服务端实例检测到的成功率
27 service_instance_relation_client_resp_time = from(ServiceInstanceRelation.latency).filter(detectPoint == DetectPoint.CLIENT).longAvg();//在客户端实例检测到的平均响应时间
28 service_instance_relation_server_resp_time = from(ServiceInstanceRelation.latency).filter(detectPoint == DetectPoint.SERVER).longAvg();//在服务端实例检测到的平均响应时间
29 service_instance_relation_client_percentile = from(ServiceInstanceRelation.latency).filter(detectPoint == DetectPoint.CLIENT).percentile(10); // Multiple values including p50, p75, p90, p95, p99
30 service_instance_relation_server_percentile = from(ServiceInstanceRelation.latency).filter(detectPoint == DetectPoint.SERVER).percentile(10); // Multiple values including p50, p75, p90, p95, p99
31 
32 // Service Instance Scope metrics
33 service_instance_sla = from(ServiceInstance.*).percent(status == true);//服务实例的成功率
34 service_instance_resp_time= from(ServiceInstance.latency).longAvg();//服务实例的平均响应时间
35 service_instance_cpm = from(ServiceInstance.*).cpm();//服务实例的每分钟调用次数
36 
37 // Endpoint scope metrics
38 endpoint_cpm = from(Endpoint.*).cpm();//端点的每分钟调用次数
39 endpoint_avg = from(Endpoint.latency).longAvg();//端口平均响应时间
40 endpoint_sla = from(Endpoint.*).percent(status == true);//端点的成功率
41 endpoint_percentile = from(Endpoint.latency).percentile(10); // Multiple values including p50, p75, p90, p95, p99
42 
43 // Endpoint relation scope metrics
44 endpoint_relation_cpm = from(EndpointRelation.*).filter(detectPoint == DetectPoint.SERVER).cpm();//在服务端端点检测到的每分钟调用次数
45 endpoint_relation_resp_time = from(EndpointRelation.rpcLatency).filter(detectPoint == DetectPoint.SERVER).longAvg();//在服务端检测到的rpc调用的平均耗时
46 endpoint_relation_sla = from(EndpointRelation.*).filter(detectPoint == DetectPoint.SERVER).percent(status == true);//在服务端检测到的请求成功率
47 endpoint_relation_percentile = from(EndpointRelation.rpcLatency).filter(detectPoint == DetectPoint.SERVER).percentile(10); // Multiple values including p50, p75, p90, p95, p99
48 
49 database_access_resp_time = from(DatabaseAccess.latency).longAvg();//数据库的处理平均响应时间
50 database_access_sla = from(DatabaseAccess.*).percent(status == true);//数据库的请求成功率
51 database_access_cpm = from(DatabaseAccess.*).cpm();//数据库的每分钟调用次数
52 database_access_percentile = from(DatabaseAccess.latency).percentile(10);

java-agent.oal

// JVM instance metrics
instance_jvm_cpu = from(ServiceInstanceJVMCPU.usePercent).doubleAvg();//jvm 平均cpu耗时百分比
instance_jvm_memory_heap = from(ServiceInstanceJVMMemory.used).filter(heapStatus == true).longAvg();//jvm 堆空间的平均使用空间
instance_jvm_memory_noheap = from(ServiceInstanceJVMMemory.used).filter(heapStatus == false).longAvg();//jvm 非堆空间的平均使用空间
instance_jvm_memory_heap_max = from(ServiceInstanceJVMMemory.max).filter(heapStatus == true).longAvg();//jvm 最大堆内存的平均值
instance_jvm_memory_noheap_max = from(ServiceInstanceJVMMemory.max).filter(heapStatus == false).longAvg();//jvm 最大非堆内存的平均值
instance_jvm_young_gc_time = from(ServiceInstanceJVMGC.time).filter(phrase == GCPhrase.NEW).sum();//年轻代gc的耗时
instance_jvm_old_gc_time = from(ServiceInstanceJVMGC.time).filter(phrase == GCPhrase.OLD).sum();//老年代gc的耗时
instance_jvm_young_gc_count = from(ServiceInstanceJVMGC.count).filter(phrase == GCPhrase.NEW).sum();//年轻代gc的次数
instance_jvm_old_gc_count = from(ServiceInstanceJVMGC.count).filter(phrase == GCPhrase.OLD).sum();//老年代gc的次数
instance_jvm_thread_live_count = from(ServiceInstanceJVMThread.liveCount).longAvg();//存活的线程数
instance_jvm_thread_daemon_count = from(ServiceInstanceJVMThread.daemonCount).longAvg();//守护线程数
instance_jvm_thread_peak_count = from(ServiceInstanceJVMThread.peakCount).longAvg();//峰值线程数

告警的设置

rules:
    # 告警规则 名称唯一 必须以_rule 结尾
  service_resp_time_rule:
      # 度量名称，只支持int long double
    metrics-name: service_resp_time
    # 操作符
    op: ">"
    # 阈值 ms
    threshold: 1000
    # 评估度量的时间长度
    period: 10
    # 度量有多少次符合告警条件后，才会触发告警
    count: 2
    # 静默时间 默认情况下，它和周期一样，在同一个周期内只会触发一次。
    silence-period: 10
    message: 服务【{name}】的平均响应时间在最近10分钟内有2分钟超过1秒
  service_sla_rule:
    metrics-name: service_sla
    op: "<"
    threshold: 8000
    period: 10
    count: 2
    silence-period: 10
    message: 服务【{name}】的成功率在最近10分钟内有2分钟低于80％
composite-rules:
  # 规则名称：在告警信息中显示的唯一名称，必须以_rule结尾
  comp_rule:
    # 指定如何组成规则，支持&&, ||, ()操作符
    expression: service_resp_time_rule && service_sla_rule
    message: 服务【{name}】在最近10分钟内有2分钟平均响应时间超过1秒并且成功率低于80％

本文章为转载内容，我们尊重原作者对文章享有的著作权。如有内容错误或侵权问题，欢迎原作者联系我们进行内容更正或删除文章。