hectorzhao/lislgosms
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix: align reporting fields queries and disk monitoring

Browse Source
This commit is contained in:
hectorzhao
2026-08-31 12:19:39 +08:00
parent 119d57772e
commit b0deef5e6e
37 changed files with 596 additions and 110 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/monitoring/cmpp-alerts.yml
+12 -12
View File
@@ -108,50 +108,50 @@ groups:
threshold: "95%"
- alert: HostRootDiskUsageWarning
expr: ((1 - node_filesystem_avail_bytes{mountpoint="/",fstype!~"tmpfs|overlay|squashfs"} / node_filesystem_size_bytes{mountpoint="/",fstype!~"tmpfs|overlay|squashfs"}) * 100 > 80) and ((1 - node_filesystem_avail_bytes{mountpoint="/",fstype!~"tmpfs|overlay|squashfs"} / node_filesystem_size_bytes{mountpoint="/",fstype!~"tmpfs|overlay|squashfs"}) * 100 <= 90)
expr: ((1 - node_filesystem_avail_bytes{device=~"/dev/.+",fstype!~"tmpfs|devtmpfs|overlay|squashfs|ramfs"} / node_filesystem_size_bytes{device=~"/dev/.+",fstype!~"tmpfs|devtmpfs|overlay|squashfs|ramfs"}) * 100 > 80) and ((1 - node_filesystem_avail_bytes{device=~"/dev/.+",fstype!~"tmpfs|devtmpfs|overlay|squashfs|ramfs"} / node_filesystem_size_bytes{device=~"/dev/.+",fstype!~"tmpfs|devtmpfs|overlay|squashfs|ramfs"}) * 100 <= 90)
for: 15m
labels:
severity: warning
service: host
annotations:
summary: 根文件系统空间不足
description: 根文件系统使用率连续15分钟高于80%。
summary: 磁盘文件系统空间不足
description: "磁盘文件系统使用率连续15分钟高于80%。 挂载点：{{ $labels.mountpoint }}；设备：{{ $labels.device }}。"
currentValue: "{{ printf \"%.1f\" $value }}%"
threshold: "80%"
- alert: HostRootDiskUsageCritical
expr: (1 - node_filesystem_avail_bytes{mountpoint="/",fstype!~"tmpfs|overlay|squashfs"} / node_filesystem_size_bytes{mountpoint="/",fstype!~"tmpfs|overlay|squashfs"}) * 100 > 90
expr: (1 - node_filesystem_avail_bytes{device=~"/dev/.+",fstype!~"tmpfs|devtmpfs|overlay|squashfs|ramfs"} / node_filesystem_size_bytes{device=~"/dev/.+",fstype!~"tmpfs|devtmpfs|overlay|squashfs|ramfs"}) * 100 > 90
for: 5m
labels:
severity: critical
service: host
annotations:
summary: 根文件系统空间严重不足
description: 根文件系统使用率连续5分钟高于90%。
summary: 磁盘文件系统空间严重不足
description: "磁盘文件系统使用率连续5分钟高于90%。 挂载点：{{ $labels.mountpoint }}；设备：{{ $labels.device }}。"
currentValue: "{{ printf \"%.1f\" $value }}%"
threshold: "90%"
- alert: HostRootInodeUsageWarning
expr: ((1 - node_filesystem_files_free{mountpoint="/",fstype!~"tmpfs|overlay|squashfs"} / node_filesystem_files{mountpoint="/",fstype!~"tmpfs|overlay|squashfs"}) * 100 > 80) and ((1 - node_filesystem_files_free{mountpoint="/",fstype!~"tmpfs|overlay|squashfs"} / node_filesystem_files{mountpoint="/",fstype!~"tmpfs|overlay|squashfs"}) * 100 <= 90)
expr: ((1 - node_filesystem_files_free{device=~"/dev/.+",fstype!~"tmpfs|devtmpfs|overlay|squashfs|ramfs"} / node_filesystem_files{device=~"/dev/.+",fstype!~"tmpfs|devtmpfs|overlay|squashfs|ramfs"}) * 100 > 80) and ((1 - node_filesystem_files_free{device=~"/dev/.+",fstype!~"tmpfs|devtmpfs|overlay|squashfs|ramfs"} / node_filesystem_files{device=~"/dev/.+",fstype!~"tmpfs|devtmpfs|overlay|squashfs|ramfs"}) * 100 <= 90)
for: 15m
labels:
severity: warning
service: host
annotations:
summary: 根文件系统inode余量偏低
description: 根文件系统inode使用率连续15分钟高于80%。
summary: 磁盘文件系统inode余量偏低
description: "磁盘文件系统inode使用率连续15分钟高于80%。 挂载点：{{ $labels.mountpoint }}；设备：{{ $labels.device }}。"
currentValue: "{{ printf \"%.1f\" $value }}%"
threshold: "80%"
- alert: HostRootInodeUsageCritical
expr: (1 - node_filesystem_files_free{mountpoint="/",fstype!~"tmpfs|overlay|squashfs"} / node_filesystem_files{mountpoint="/",fstype!~"tmpfs|overlay|squashfs"}) * 100 > 90
expr: (1 - node_filesystem_files_free{device=~"/dev/.+",fstype!~"tmpfs|devtmpfs|overlay|squashfs|ramfs"} / node_filesystem_files{device=~"/dev/.+",fstype!~"tmpfs|devtmpfs|overlay|squashfs|ramfs"}) * 100 > 90
for: 5m
labels:
severity: critical
service: host
annotations:
summary: 根文件系统inode严重不足
description: 根文件系统inode使用率连续5分钟高于90%。
summary: 磁盘文件系统inode严重不足
description: "磁盘文件系统inode使用率连续5分钟高于90%。 挂载点：{{ $labels.mountpoint }}；设备：{{ $labels.device }}。"
currentValue: "{{ printf \"%.1f\" $value }}%"
threshold: "90%"