node.js - 小型 Node.js 应用程序 Pod 的 GKE CPU 不足

Question

所以在 GKE 上，我有一个 Node.js app，每个 pod 使用大约:CPU(cores): 5m, MEMORY: 100Mi

但是我只能为每个 Node 部署 1 个 pod。我使用的是 GKE n1-standard-1 集群，每个 Node 有 1 个 vCPU，3.75 GB。

所以为了获得 2 个 app pod 总计 = CPU(cores): 10m, MEMORY: 200Mi, 它需要另一个完整的 +1 Node = 2 个 Node = 2 vCPU, 7.5 GB 使其工作。如果我尝试将这 2 个 pod 部署在同一个 Node 上，我会收到 insufficient CPU 错误。

我有一种感觉，我实际上应该能够在 f1-micro 的 1 个 Node (1 个 vCPU，0.6 GB)或 f1-small(1 个 vCPU，1.7 GB)，我在这里过度配置，浪费了我的钱。

但我不确定为什么我似乎受到 CPU 不足 的限制。我需要更改一些配置吗？任何指导将不胜感激。

---

Allocatable:
 cpu:                940m
 ephemeral-storage:  47093746742
 hugepages-2Mi:      0
 memory:             2702216Ki
 pods:               110
Non-terminated Pods:         (7 in total)
  Namespace                  Name                                                CPU Requests  CPU Limits  Memory Requests  Memory Limits
  ---------                  ----                                                ------------  ----------  ---------------  -------------
  default                    mission-worker-5cf6654687-fwmk4                     100m (10%)    0 (0%)      0 (0%)           0 (0%)
  default                    mission-worker-5cf6654687-lnwkt                     100m (10%)    0 (0%)      0 (0%)           0 (0%)
  kube-system                fluentd-gcp-v3.1.1-5b6km                            100m (10%)    1 (106%)    200Mi (7%)       500Mi (18%)
  kube-system                kube-dns-76dbb796c5-jgljr                           260m (27%)    0 (0%)      110Mi (4%)       170Mi (6%)
  kube-system                kube-proxy-gke-test-cluster-pool-1-96c6d8b2-m15p    100m (10%)    0 (0%)      0 (0%)           0 (0%)
  kube-system                metadata-agent-nb4dp                                40m (4%)      0 (0%)      50Mi (1%)        0 (0%)
  kube-system                prometheus-to-sd-gwlkv                              1m (0%)       3m (0%)     20Mi (0%)        20Mi (0%)
Allocated resources:
  (Total limits may be over 100 percent, i.e., overcommitted.)
  Resource  Requests     Limits
  --------  --------     ------
  cpu       701m (74%)   1003m (106%)
  memory    380Mi (14%)  690Mi (26%)
Events:     <none>

Answer 1

部署后，使用kubectl describe nodes检查 Node 容量。例如:在答案底部的代码示例中:

可分配 cpu:1800m

已被 kube-system 命名空间中的 pod 使用:100m + 260m++100m + 200m + 20m = 680m

也就是说1800m - 680m = 1120m留给你使用

因此，如果您的一个或多个 pod 请求超过 1120m 的 cpu，它们将不适合此 Node

So in order to get 2 pods of app up total = CPU(cores): 10m, MEMORY: 200Mi, it requires another entire +1 node = 2 nodes = 2 vCPU, 7.5 GB to make it work. If I try to deploy those 2 pods on the same single node, I get insufficient CPU error.

如果您进行上述练习，您会找到答案。如果有足够的 cpu 供您的 pod 使用，但您仍然收到 CPU 不足的错误，请检查您是否正确设置了 cpu 请求和限制参数。参见 here

如果您执行了上述所有操作，但问题仍然存在。然后，我认为在您的情况下，可能发生的情况是您正在为您的 Node 应用程序分配 5-10m cpu，这太少了 cpu 无法分配。尝试将其增加到 50m cpu。

I have a feeling I should actually be able to run a handful of pod replicas (like 3 replicas and more) on 1 node of f1-micro (1 vCPU, 0.6 GB) or f1-small (1 vCPU, 1.7 GB), and that I am way overprovisioned here, and wasting my money.

再一次，做上面描述的练习以得出结论

Name:            e2e-test-minion-group-4lw4
[ ... lines removed for clarity ...]
Capacity:
 cpu:                               2
 memory:                            7679792Ki
 pods:                              110
Allocatable:
 cpu:                               1800m
 memory:                            7474992Ki
 pods:                              110
[ ... lines removed for clarity ...]
Non-terminated Pods:        (5 in total)
  Namespace    Name                                  CPU Requests  CPU Limits  Memory Requests  Memory Limits
  ---------    ----                                  ------------  ----------  ---------------  -------------
  kube-system  fluentd-gcp-v1.38-28bv1               100m (5%)     0 (0%)      200Mi (2%)       200Mi (2%)
  kube-system  kube-dns-3297075139-61lj3             260m (13%)    0 (0%)      100Mi (1%)       170Mi (2%)
  kube-system  kube-proxy-e2e-test-...               100m (5%)     0 (0%)      0 (0%)           0 (0%)
  kube-system  monitoring-influxdb-grafana-v4-z1m12  200m (10%)    200m (10%)  600Mi (8%)       600Mi (8%)
  kube-system  node-problem-detector-v0.1-fj7m3      20m (1%)      200m (10%)  20Mi (0%)        100Mi (1%)
Allocated resources:
  (Total limits may be over 100 percent, i.e., overcommitted.)
  CPU Requests    CPU Limits    Memory Requests    Memory Limits
  ------------    ----------    ---------------    -------------
  680m (34%)      400m (20%)    920Mi (12%)        1070Mi (14%)