【Prometheus】Metric类型的理解及运用

ゝ一纸荒年。 2023-06-28 06:16 42阅读 0赞

### Metric类型 ###

**1.Counter**  
**作用:** 累加的计数器

`样本样例:`

#HELP node_cpu Seconds the cpus spent in each mode.
    #TYPE node_cpu counter
    node_cpu{cpu="cpu0",mode="idle"} 362812.7890625

eg1: 计算5分钟内http请求量增长率

rate(http_requests_total[5m])

eg2: 查询请求量前10的http地址

topk(10, http_requests_total)

--------------------

**2.Gauge**  
**作用:** 实时反应当前状态的仪表盘  
eg1: 计算php容器内存消耗在2小时内差异

delta(container_memory_usage_bytes{ container='php'}[2h])

eg2: 计算上述样本的线性回归模型

deriv(container_memory_usage_bytes{ container='php'}[2h])

**经典样本计算:**  
`样本样例：`

container_memory_working_set_bytes{ container="POD",container_name="POD",image="k8s.gcr.io/pause:3.1",instance="k8s-node2",job="kubernetes-cadvisor",kubernetes_io_hostname="k8s-node2",name="k8s_POD_kube-proxy-6ppd2_kube-system_9f12d89b-2ae6-11ea-b5dc-fa163eb7be75_14",namespace="kube-system",pod="kube-proxy-6ppd2",pod_name="kube-proxy-6ppd2"}

eg1:查询实时Pod内存使用量

sum(container_memory_working_set_bytes{ image!=''}) by (pod_name)

sum的意思把某一时刻下所符合的列表值累加，by的意思是通过pod\_name分组  
整体的意思是：实时绘制各个pod的内存使用状况

**那为什么要用sum呢？**  
其实k8s中部署Pod后，不管你的pod内有多少容器，都会给你额外匹配pause容器，这个容器的作用相当于Pod内的基础环境，构建Pod内容器的公共设施(命名空间，网络，IPC等)，可类比虚拟机。  
所以统计Pod内存使用量就需要把所有pod下所有容器值相加  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4OTAwNTY1_size_16_color_FFFFFF_t_70]  
反之，不用sum的话就会查询到所有容器的内存使用量了

**举一反三:**  
eg1.那么如何计算各节点下的容器使用量总值？  
首先是通过节点分组:by (instance)  
然后就是组下的各容器内存值累加

sum(container_memory_working_set_bytes{ image!=''}) by (instance)

eg2.根据1小时前的情况，预测接下来2小时的各节点内存使用量状况

sum(predict_linear(container_memory_working_set_bytes{ image!=''}[1h],2*3600)) by (instance)

--------------------

**3.Histogram & Summary**  
**作用:** 统计、分析数据分布情况  
例如了解请求耗时分布情况

**3.1 Histogram样例**

# HELP coredns_forward_request_duration_seconds Histogram of the time each request took.
    # TYPE coredns_forward_request_duration_seconds histogram
    coredns_forward_request_duration_seconds_bucket{
       to="114.114.114.114:53",le="0.00025"} 0.0
    coredns_forward_request_duration_seconds_bucket{
       to="114.114.114.114:53",le="0.0005"} 0.0
    coredns_forward_request_duration_seconds_bucket{
       to="114.114.114.114:53",le="0.001"} 0.0
    coredns_forward_request_duration_seconds_bucket{
       to="114.114.114.114:53",le="0.002"} 0.0
    coredns_forward_request_duration_seconds_bucket{
       to="114.114.114.114:53",le="0.004"} 0.0
    coredns_forward_request_duration_seconds_bucket{
       to="114.114.114.114:53",le="0.008"} 0.0
    coredns_forward_request_duration_seconds_bucket{
       to="114.114.114.114:53",le="0.016"} 0.0
    coredns_forward_request_duration_seconds_bucket{
       to="114.114.114.114:53",le="0.032"} 259.0
    coredns_forward_request_duration_seconds_bucket{
       to="114.114.114.114:53",le="0.064"} 327.0
    coredns_forward_request_duration_seconds_bucket{
       to="114.114.114.114:53",le="0.128"} 344.0
    coredns_forward_request_duration_seconds_bucket{
       to="114.114.114.114:53",le="0.256"} 351.0
    coredns_forward_request_duration_seconds_bucket{
       to="114.114.114.114:53",le="0.512"} 355.0
    coredns_forward_request_duration_seconds_bucket{
       to="114.114.114.114:53",le="1.024"} 356.0
    coredns_forward_request_duration_seconds_bucket{
       to="114.114.114.114:53",le="2.048"} 357.0
    coredns_forward_request_duration_seconds_bucket{
       to="114.114.114.114:53",le="4.096"} 357.0
    coredns_forward_request_duration_seconds_bucket{
       to="114.114.114.114:53",le="8.192"} 357.0
    coredns_forward_request_duration_seconds_bucket{
       to="114.114.114.114:53",le="+Inf"} 357.0
    coredns_forward_request_duration_seconds_sum{
       to="114.114.114.114:53"} 15.852329659999997
    coredns_forward_request_duration_seconds_count{
       to="114.114.114.114:53"} 357.0

该指标表示K8s内dns转发请求消耗时间  
总操作数为357次  
总持续时间为15s  
其中耗时在0.128内(le=0.128)的操作次数为344次  
在耗时(128,256\] 区间的操作次数为351-344=11次

**3.2 Summary样例**

# HELP go_gc_duration_seconds A summary of the GC invocation durations.
    # TYPE go_gc_duration_seconds summary
    go_gc_duration_seconds{quantile="0.0"} 3.2722e-05
    go_gc_duration_seconds{quantile="0.25"} 8.0476e-05
    go_gc_duration_seconds{quantile="0.5"} 9.6085e-05
    go_gc_duration_seconds{quantile="0.75"} 0.00012454
    go_gc_duration_seconds{quantile="1.0"} 0.003750961
    go_gc_duration_seconds_sum 1.228575062
    go_gc_duration_seconds_count 6437.0

该指标表示go的gc垃圾回收持续时间  
总操作数为6437次  
总持续时间为1.228575062s  
其中中位数(quantile=0.5)的持续时间为9.6085e-05s

**3.3 差异**  
**3.3.1** Histogram直接反应了在不同区间内样本的个数，区间通过标签len进行定义;  
**3.3.2** Histogram可以通过histogram\_quantile()函数计算出其值的分位数。但是在服务器端进行计算，而summary则是在客户端计算完成。

[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4OTAwNTY1_size_16_color_FFFFFF_t_70]: https://img-blog.csdnimg.cn/20200107190134371.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4OTAwNTY1,size_16,color_FFFFFF,t_70