node-problem-detector-configmap.yaml 2.9 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081
  1. apiVersion: v1
  2. kind: ConfigMap
  3. metadata:
  4. name: npd-config
  5. data:
  6. docker-monitor.json: |
  7. {
  8. "plugin": "journald",
  9. "pluginConfig": {
  10. "source": "docker"
  11. },
  12. "logPath": "/host/log/journal",
  13. "lookback": "5m",
  14. "bufferSize": 10,
  15. "source": "docker-monitor",
  16. "conditions": [],
  17. "rules": [
  18. {
  19. "type": "temporary",
  20. "reason": "CorruptDockerImage",
  21. "pattern": "Error trying v2 registry: failed to register layer: rename /var/lib/docker/image/(.+) /var/lib/docker/image/(.+): directory not empty.*"
  22. }
  23. ]
  24. }
  25. kernel-monitor.json: |
  26. {
  27. "plugin": "journald",
  28. "pluginConfig": {
  29. "source": "kernel"
  30. },
  31. "logPath": "/host/log/journal",
  32. "lookback": "5m",
  33. "bufferSize": 10,
  34. "source": "kernel-monitor",
  35. "conditions": [
  36. {
  37. "type": "KernelDeadlock",
  38. "reason": "KernelHasNoDeadlock",
  39. "message": "kernel has no deadlock"
  40. }
  41. ],
  42. "rules": [
  43. {
  44. "type": "temporary",
  45. "reason": "OOMKilling",
  46. "pattern": "Kill process \\d+ (.+) score \\d+ or sacrifice child\\nKilled process \\d+ (.+) total-vm:\\d+kB, anon-rss:\\d+kB, file-rss:\\d+kB"
  47. },
  48. {
  49. "type": "temporary",
  50. "reason": "TaskHung",
  51. "pattern": "task \\S+:\\w+ blocked for more than \\w+ seconds\\."
  52. },
  53. {
  54. "type": "temporary",
  55. "reason": "UnregisterNetDevice",
  56. "pattern": "unregister_netdevice: waiting for \\w+ to become free. Usage count = \\d+"
  57. },
  58. {
  59. "type": "temporary",
  60. "reason": "KernelOops",
  61. "pattern": "BUG: unable to handle kernel NULL pointer dereference at .*"
  62. },
  63. {
  64. "type": "temporary",
  65. "reason": "KernelOops",
  66. "pattern": "divide error: 0000 \\[#\\d+\\] SMP"
  67. },
  68. {
  69. "type": "permanent",
  70. "condition": "KernelDeadlock",
  71. "reason": "AUFSUmountHung",
  72. "pattern": "task umount\\.aufs:\\w+ blocked for more than \\w+ seconds\\."
  73. },
  74. {
  75. "type": "permanent",
  76. "condition": "KernelDeadlock",
  77. "reason": "DockerHung",
  78. "pattern": "task docker:\\w+ blocked for more than \\w+ seconds\\."
  79. }
  80. ]
  81. }