当前位置: 首页 > news >正文

AI编程时代,人类程序员还剩下什么?蒙

故障表现

发现请求集群 demo 入口时卡住,并且对应 Pod 没有新的日志输出

root@ce-demo-1:~# kubectl get pods -n deepflow-otel-spring-demo -o wide

NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES

db-demo-0 1/1 Running 0 2d1h 10.244.81.203 ce-demo-3

nacos-0 1/1 Running 0 2d1h 10.244.142.37 ce-demo-1

svc-item-588b4cfcc9-zkvpd 1/1 Running 0 2d1h 10.244.142.36 ce-demo-1

svc-order-5f67c67555-ds7bj 1/1 Running 0 2d1h 10.244.228.140 ce-demo-2

svc-stock-b9df64d4b-bsxs5 1/1 Running 0 2d1h 10.244.142.38 ce-demo-1

svc-user-7c5c7b488f-4zjdc 1/1 Running 0 2d1h 10.244.81.204 ce-demo-3

web-shop-5d495d8cbc-lnpxq 1/1 Running 0 2d1h 10.244.228.139 ce-demo-2

root@ce-demo-1:~# curl 10.244.228.139:8090/shop/full-test

## 等待很久后报错

curl: (28) Failed to connect to 10.244.228.139 port 8090 after 133345 ms: Could not connect to server

## 此处新开终端查看日志后再次请求,发现没有新的日志

root@ce-demo-1:~# kubectl logs -f -n deepflow-otel-spring-demo web-shop-5d495d8cbc-lnpxq --tail 20

排查流程

在客户端抓包发现请求后一直建连失败,从客户端 IP 看出,识别到 Server IP 是 Pod IP,所以直接走了 calico 网口转发

image

tcpdump -v -i any dst 10.244.228.139 -w pod.pcap

root@ce-demo-1:~# ip a s vxlan.calico

30499: vxlan.calico: mtu 1450 qdisc noqueue state UNKNOWN group default qlen 1000

link/ether 66:e0:bb:93:52:4f brd ff:ff:ff:ff:ff:ff

inet 10.244.142.0/32 scope global vxlan.calico

valid_lft forever preferred_lft forever

由上数据怀疑是 ce-demo-2 节点 calico 问题,查看后发现对应节点 calico-node 运行异常:

此组件作用可参考官网 calico/node 配置说明 和 calico 组件架构

root@ce-demo-1:~# kubectl get pods -n calico-system -o wide

NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES

calico-kube-controllers-86596856c4-w7nsl 1/1 Running 0 2d9h 10.244.228.132 ce-demo-2

calico-node-n2b4b 1/1 Running 0 2d9h 10.51.0.102 ce-demo-3

calico-node-p4k7s 1/1 Running 0 2d9h 10.51.0.100 ce-demo-1

calico-node-sbxrk 0/1 Running 0 44h 10.51.0.101 ce-demo-2

calico-typha-75c74d6ffd-6xq2j 1/1 Running 0 2d9h 10.51.0.100 ce-demo-1

calico-typha-75c74d6ffd-fkf6b 1/1 Running 0 2d9h 10.51.0.102 ce-demo-3

csi-node-driver-72pkg 2/2 Running 0 2d9h 10.244.142.1 ce-demo-1

csi-node-driver-8sjvc 2/2 Running 0 2d9h 10.244.81.195 ce-demo-3

csi-node-driver-skj6v 2/2 Running 0 2d9h 10.244.228.130 ce-demo-2

Calico 启动 calico-node 时,会根据主机的网络接口自动选择一个 IP 地址作为节点的 IPv4Address,默认是自动探测(autodetect)。查看三台节点对应的值,发现此 ip 不是对应节点的 ens160:

root@ce-demo-1:~# kubectl get node -o yaml | grep 'IPv4Address'

projectcalico.org/IPv4Address: 10.51.0.100/24

projectcalico.org/IPv4Address: 10.4.0.1/24

projectcalico.org/IPv4Address: 10.51.0.102/24

ce-demo-2 节点查看 10.1.0.1 对应 ip 网络设备的详细信息,可以看出这是一个网桥(bridge)设备,用于容器间通信/外部访问

root@ce-demo-2:~# ip address show nerdctl0

686: nerdctl0: mtu 1500 qdisc noqueue state UP group default qlen 1000

link/ether b2:90:e4:5b:f4:d1 brd ff:ff:ff:ff:ff:ff

inet 10.4.0.1/24 brd 10.4.0.255 scope global nerdctl0

valid_lft forever preferred_lft forever

root@ce-demo-2:~# ip -d link show nerdctl0

686: nerdctl0: mtu 1500 qdisc noqueue state UP mode DEFAULT group default qlen 1000

link/ether b2:90:e4:5b:f4:d1 brd ff:ff:ff:ff:ff:ff promiscuity 0 allmulti 0 minmtu 68 maxmtu 65535

bridge forward_delay 1500 hello_time 200 max_age 2000 ageing_time 30000 stp_state 0 priority 32768 vlan_filtering 0 vlan_protocol 802.1Q bridge_id 8000.b2:90:e4:5b:f4:d1 designated_root 8000.b2:90:e4:5b:f4:d1 root_port 0 root_path_cost 0 topology_change 0 topology_change_detected 0 hello_timer 0.00 tcn_timer 0.00 topology_change_timer 0.00 gc_timer 240.02 vlan_default_pvid 1 vlan_stats_enabled 0 vlan_stats_per_port 0 group_fwd_mask 0 group_address 01:80:c2:00:00:00 mcast_snooping 1 no_linklocal_learn 0 mcast_vlan_snooping 0 mcast_router 1 mcast_query_use_ifaddr 0 mcast_querier 0 mcast_hash_elasticity 16 mcast_hash_max 4096 mcast_last_member_count 2 mcast_startup_query_count 2 mcast_last_member_interval 100 mcast_membership_interval 26000 mcast_querier_interval 25500 mcast_query_interval 12500 mcast_query_response_interval 1000 mcast_startup_query_interval 3125 mcast_stats_enabled 0 mcast_igmp_version 2 mcast_mld_version 1 nf_call_iptables 0 nf_call_ip6tables 0 nf_call_arptables 0 addrgenmode eui64 numtxqueues 1 numrxqueues 1 gso_max_size 65536 gso_max_segs 65535 tso_max_size 524280 tso_max_segs 65535 gro_max_size 65536

root@ce-demo-2:~# brctl show nerdctl0

bridge name bridge id STP enabled interfaces

nerdctl0 8000.b290e45bf4d1 no veth76f5219b

Calico 默认配置通过 first-found 模式(从所有接口中第一个有 IP 的接口选一个)

从 ce-demo-2 节点 calico-node log time 看出,首先看到的是 nerdctl0:

接口遍历顺序由宿主机内核/Netlink 返回的接口顺序决定,常见情况下与接口 ifindex(创建顺序/系统里登记的编号)相关;如果 nerdctl0 在系统里排序靠前(例如创建得更早,或者 ifindex 更靠前),first-found 就会先看到它。

root@ce-demo-1:~# kubectl describe daemonset calico-node -n calico-system | grep "IP_AUTODETECTION_METHOD"

IP_AUTODETECTION_METHOD: first-found

root@ce-demo-1:~# kubectl logs -n calico-system calico-node-sbxrk -c calico-node | grep -i 'nerdctl0'

2025-07-15 07:51:58.079 [INFO][9] startup/autodetection_methods.go 103: Using autodetected IPv4 address on interface nerdctl0: 10.4.0.1/24

2025-07-15 07:52:02.351 [INFO][87] felix/int_dataplane.go 1431: Linux interface state changed. ifIndex=686 ifaceName="nerdctl0" state="up"

2025-07-15 07:52:02.351 [INFO][87] felix/int_dataplane.go 1475: Linux interface addrs changed. addrs=set.Set{10.4.0.1} ifaceName="nerdctl0"

2025-07-15 07:52:02.351 [INFO][87] felix/int_dataplane.go 2098: Received interface update msg=&intdataplane.ifaceStateUpdate{Name:"nerdctl0", State:"up", Index:686}

2025-07-15 07:52:02.351 [INFO][87] felix/int_dataplane.go 2125: Received interface addresses update msg=&intdataplane.ifaceAddrsUpdate{Name:"nerdctl0", Addrs:set.Typed[string]{"10.4.0.1":set.v{}}}

2025-07-15 07:52:02.351 [INFO][87] felix/hostip_mgr.go 84: Interface addrs changed. update=&intdataplane.ifaceAddrsUpdate{Name:"nerdctl0", Addrs:set.Typed[string]{"10.4.0.1":set.v{}}}

2025-07-15 07:52:02.413 [INFO][87] felix/vxlan_mgr.go 597: VXLAN device parent changed from "" to "nerdctl0" ipVersion=0x4

root@ce-demo-1:~# kubectl logs -n calico-system calico-node-sbxrk -c calico-node | grep -i 'ens160'

2025-07-15 07:52:02.348 [INFO][87] felix/int_dataplane.go 1431: Linux interface state changed. ifIndex=2 ifaceName="ens160" state="up"

2025-07-15 07:52:02.348 [INFO][87] felix/int_dataplane.go 1475: Linux interface addrs changed. addrs=set.Set{10.51.0.101} ifaceName="ens160"

2025-07-15 07:52:02.349 [INFO][87] felix/int_dataplane.go 2098: Received interface update msg=&intdataplane.ifaceStateUpdate{Name:"ens160", State:"up", Index:2}

2025-07-15 07:52:02.349 [INFO][87] felix/int_dataplane.go 2125: Received interface addresses update msg=&intdataplane.ifaceAddrsUpdate{Name:"ens160", Addrs:set.Typed[string]{"10.51.0.101":set.v{}}}

2025-07-15 07:52:02.349 [INFO ][87] felix/hostip_mgr.go 84: Interface addrs changed. update=&intdataplane.ifaceAddrsUpdate{Name:"ens160", Addrs:set.Typed[string]{"10.51.0.101":set.v{}}}

2025-07-16 06:17:42.649 [INFO][87] felix/int_dataplane.go 1475: Linux interface addrs changed. addrs=set.Set{10.51.0.101,fe80::20c:29ff:febb:1bdc} ifaceName="ens160"

2025-07-16 06:17:42.649 [INFO][87] felix/int_dataplane.go 2125: Received interface addresses update msg=&intdataplane.ifaceAddrsUpdate{Name:"ens160", Addrs:set.Typed[string]{"10.51.0.101":set.v{}, "fe80::20c:29ff:febb:1bdc":set.v{}}}

2025-07-16 06:17:42.649 [INFO][87] felix/hostip_mgr.go 84: Interface addrs changed. update=&intdataplane.ifaceAddrsUpdate{Name:"ens160", Addrs:set.Typed[string]{"10.51.0.101":set.v{}, "fe80::20c:29ff:febb:1bdc":set.v{}}}

由于 Calico 官方文档中也有注明,first-found 模式为:使用第一个接口(不包括 Docker 网桥等本地接口)上的第一个有效 IP 地址,建议根据不同需求选择不同配置方式,本文采用的指定接口方式。

还需要额外解释下,first-found 文档中描述的 "the first valid IP address on the first interface (excluding local interfaces such as the docker bridge)." 只是举例,他的默认排除项中并不包括 nerdctl0,所以它会被当作一个合法候选接口。

## 注:需要看 Calico 部署方式,当前环境通过 Calico Tigera Operator 自定义资源部署,无法直接更改 DaemonSet

root@ce-demo-1:~# kubectl edit daemonset calico-node -n calico-system

## 找到或添加这个变量

- name: IP_AUTODETECTION_METHOD

## 由于几台机器网卡名称都是 ens160,此处写网卡名或正则都可以

value: interface=ens.*

需要更改 Calico 的自定义资源参数,官网链接搜索 nodeAddressAutodetectionV4 后有具体参数:

root@ce-demo-1:~# kubectl get Installation

NAME AGE

default 2d10h

root@ce-demo-1:~# kubectl edit installation default

nodeAddressAutodetectionV4:

## 默认使用此配置

##firstFound: true

## 删掉后添加指定接口:

interface: ens160

## 查看更新结果

root@ce-demo-1:~# kubectl get pods -n calico-system -o wide | grep calico-node

calico-node-4fndv 1/1 Running 0 104s 10.51.0.101 ce-demo-2

calico-node-8n5qr 1/1 Running 0 39s 10.51.0.100 ce-demo-1

calico-node-rxhsf 1/1 Running 0 72s 10.51.0.102 ce-demo-3

root@ce-demo-1:~# kubectl describe daemonset -n calico-system calico-node | grep "IP_AUTODETECTION_METHOD"

IP_AUTODETECTION_METHOD: interface=ens160

验证更新后效果

root@ce-demo-1:~# curl 10.244.228.139:8090/shop/full-test ; echo

{"count":1,"elapsed":209,"elapsedAvg":209,"startAt":"2025-07-17 13:33:55.116","stopAt":"2025-07-17 13:33:55.325","success"嚎庸质思

http://www.jsqmd.com/news/621119/

相关文章:

  • ESP32驱动NIDEC 24H电机控制器实战指南
  • Android 源码预创建 /data 目录的方法
  • CentOS 7.4编译FFmpeg遇阻:从nasm/yasm报错到完整安装的实战指南
  • 桌面端 Claw 个人微信接入指南炯
  • 写作柚AI——快速论文降重
  • 热源强度分布函数
  • Godot 4.5 入门教程:101. 项目准备基本窗口
  • 终极老旧Mac升级指南:OpenCore Legacy Patcher完整教程
  • CSS如何让带Flex属性的元素自身不脱离文本流控制
  • 2026黑客入门到精通必看书单!全网超全整理,一篇搞定不用再找
  • AI原生App开发不再遥不可及:2026奇点大会首发的7个可即插即用架构模板(含iOS/Android/AI芯片协同SDK)
  • BM92S2222-A指纹模块嵌入式驱动与Arduino开发指南
  • Geoserver结合WMTS实现高效瓦片地图生产全流程解析
  • 闲人闲谈PS之三十四——SAP承诺成本在项目预算控制中的实战推演
  • 戴上Rokid眼镜就能省钱!手把手教你开发AI比价助手
  • 存储那么贵,何不白嫖飞书云文件空间敝
  • mysql执行预处理语句流程是怎样的_SQL执行优化解析
  • 2026Q2四川商场酒店消防改造:合规要点与靠谱服务商指南 - 优质品牌商家
  • 如何用Bilibili-Evolved打造你的专属B站体验:5大核心功能完全指南
  • 为什么 Edge Compare 能过,Window Compare 却 fail?顺便把 93K 和 750HD 一次讲透
  • UE4 UObject垃圾回收机制深度解析:从UPROPERTY标记到内存释放
  • # 011、AutoSAR CP基础软件(BSW)模块详解:服务层
  • PS5破解新姿势:零外网依赖的DNS劫持方案(含OpenWrt路由配置)
  • 告别ArcGIS Server高成本!手把手教你用GeoServer 2.16发布ArcGIS 10.2切片包
  • 具身智能的“物理底座”:为什么说硬件工程决定了算法的生死?
  • 2026四川写字楼消防维保标杆名录:专业消防维保服务公司/写字楼消防维保公司推荐/厂房消防改造公司/选择指南 - 优质品牌商家
  • MMC-HVDC仿真模型及柔性直流输电相关基础模型集合
  • 从经典到现代:探索成核理论的演变与应用
  • 从0到1打造完美PRD:这10个细节让你的需求文档更专业
  • ESlint 规则配置与团队规范