a_zhen
/
nightingale
forked from ccfos/nightingale
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

code refactor

This commit is contained in:
Ulric Qin 2023-06-21 14:59:38 +08:00
parent 563fb0330a
commit 58136d30e6
1 changed files with 511 additions and 154 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

665
integrations/linux/alerts/linux_by_categraf.json
View File

@ -1,7 +1,141 @@
[
{
"cate": "host",
"datasource_ids": null,
"cate": "prometheus",
"datasource_ids": [
0
],
"name": "Hard disk - expected to be written full in 4 hours - categraf",
"note": "",
"prod": "metric",
"algorithm": "",
"algo_params": null,
"delay": 0,
"severity": 1,
"severities": [
1
],
"disabled": 1,
"prom_for_duration": 60,
"prom_ql": "predict_linear(disk_free[1h], 4*3600) < 0",
"rule_config": {
"algo_params": null,
"inhibit": false,
"prom_ql": "",
"queries": [
{
"prom_ql": "predict_linear(disk_free[1h], 4*3600) < 0",
"severity": 1
}
],
"severity": 0
},
"prom_eval_interval": 15,
"enable_stime": "",
"enable_stimes": [
"00:00"
],
"enable_etime": "",
"enable_etimes": [
"23:59"
],
"enable_days_of_week": null,
"enable_days_of_weeks": [
[
"1",
"2",
"3",
"4",
"5",
"6",
"0"
]
],
"enable_in_bg": 0,
"notify_recovered": 1,
"notify_channels": [
"email",
"dingtalk",
"wecom"
],
"notify_repeat_step": 60,
"notify_max_number": 0,
"recover_duration": 0,
"callbacks": [],
"runbook_url": "",
"append_tags": [],
"annotations": null
},
{
"cate": "prometheus",
"datasource_ids": [
0
],
"name": "Hard disk - IO is a bit busy - categraf",
"note": "",
"prod": "metric",
"algorithm": "",
"algo_params": null,
"delay": 0,
"severity": 2,
"severities": [
2
],
"disabled": 0,
"prom_for_duration": 60,
"prom_ql": "rate(diskio_io_time[1m])/10 > 99",
"rule_config": {
"algo_params": null,
"inhibit": false,
"prom_ql": "",
"queries": [
{
"prom_ql": "rate(diskio_io_time[1m])/10 > 99",
"severity": 2
}
],
"severity": 0
},
"prom_eval_interval": 15,
"enable_stime": "",
"enable_stimes": [
"00:00"
],
"enable_etime": "",
"enable_etimes": [
"23:59"
],
"enable_days_of_week": null,
"enable_days_of_weeks": [
[
"1",
"2",
"3",
"4",
"5",
"6",
"0"
]
],
"enable_in_bg": 0,
"notify_recovered": 1,
"notify_channels": [
"email",
"dingtalk",
"wecom"
],
"notify_repeat_step": 60,
"notify_max_number": 0,
"recover_duration": 0,
"callbacks": [],
"runbook_url": "",
"append_tags": [],
"annotations": null
},
{
"cate": "prometheus",
"datasource_ids": [
0
],
"name": "Lost connection with monitoring target - categraf",
"note": "",
"prod": "host",
@ -9,6 +143,9 @@
"algo_params": null,
"delay": 0,
"severity": 0,
"severities": [
0
],
"disabled": 0,
"prom_for_duration": 0,
"prom_ql": "",
@ -30,23 +167,15 @@
]
},
"prom_eval_interval": 15,
"enable_stime": "00:00",
"enable_stime": "",
"enable_stimes": [
"00:00"
],
"enable_etime": "23:59",
"enable_etime": "",
"enable_etimes": [
"23:59"
],
"enable_days_of_week": [
"1",
"2",
"3",
"4",
"5",
"6",
"0"
],
"enable_days_of_week": null,
"enable_days_of_weeks": [
[
"1",
@ -69,47 +198,55 @@
"append_tags": [],
"annotations": {}
},
{
"name": "Machine load - high memory, please pay attention - categraf",
"note": "",
"severity": 2,
"disabled": 0,
"prom_for_duration": 60,
"prom_ql": "mem_available_percent < 25",
"prom_eval_interval": 15,
"enable_stime": "00:00",
"enable_etime": "23:59",
"enable_days_of_week": [
"1",
"2",
"3",
"4",
"5",
"6",
"0"
],
"notify_recovered": 1,
"notify_channels": [
"email",
"dingtalk",
"wecom"
],
"notify_repeat_step": 60,
"callbacks": [],
"runbook_url": "",
"append_tags": []
{
"cate": "prometheus",
"datasource_ids": [
0
],
"name": "Machine load - high memory, please pay attention - categraf",
"note": "",
"prod": "metric",
"algorithm": "",
"algo_params": null,
"delay": 0,
"severity": 0,
"severities": [
3,
2,
1
],
"disabled": 0,
"prom_for_duration": 60,
"prom_ql": "",
"rule_config": {
"inhibit": true,
"queries": [
{
"prom_ql": "mem_available_percent < 25",
"severity": 3
},
{
"prom_ql": "mem_available_percent < 15",
"severity": 2
},
{
"prom_ql": "mem_available_percent < 5",
"severity": 1
}
]
},
{
"name": "Hard disk - IO is a bit busy - categraf",
"note": "",
"severity": 2,
"disabled": 0,
"prom_for_duration": 60,
"prom_ql": "rate(diskio_io_time[1m])/10 > 99",
"prom_eval_interval": 15,
"enable_stime": "00:00",
"enable_etime": "23:59",
"enable_days_of_week": [
"prom_eval_interval": 15,
"enable_stime": "",
"enable_stimes": [
"00:00"
],
"enable_etime": "",
"enable_etimes": [
"23:59"
],
"enable_days_of_week": null,
"enable_days_of_weeks": [
[
"1",
"2",
"3",
@ -117,29 +254,65 @@
"5",
"6",
"0"
]
],
"enable_in_bg": 0,
"notify_recovered": 1,
"notify_channels": [
"email",
"dingtalk",
"wecom"
],
"notify_repeat_step": 60,
"notify_max_number": 0,
"recover_duration": 0,
"callbacks": [],
"runbook_url": "",
"append_tags": [],
"annotations": {}
},
{
"cate": "prometheus",
"datasource_ids": [
0
],
"name": "number of TME_WAIT exceeds 20,000 - categraf",
"note": "",
"prod": "metric",
"algorithm": "",
"algo_params": null,
"delay": 0,
"severity": 2,
"severities": [
2
],
"disabled": 0,
"prom_for_duration": 60,
"prom_ql": "netstat_tcp_time_wait > 20000",
"rule_config": {
"algo_params": null,
"inhibit": false,
"prom_ql": "",
"queries": [
{
"prom_ql": "netstat_tcp_time_wait > 20000",
"severity": 2
}
],
"notify_recovered": 1,
"notify_channels": [
"email",
"dingtalk",
"wecom"
],
"notify_repeat_step": 60,
"callbacks": [],
"runbook_url": "",
"append_tags": []
"severity": 0
},
{
"name": "Hard disk - expected to be written full in 4 hours - categraf",
"note": "",
"severity": 1,
"disabled": 0,
"prom_for_duration": 60,
"prom_ql": "predict_linear(disk_free[1h], 4*3600) < 0",
"prom_eval_interval": 15,
"enable_stime": "00:00",
"enable_etime": "23:59",
"enable_days_of_week": [
"prom_eval_interval": 15,
"enable_stime": "",
"enable_stimes": [
"00:00"
],
"enable_etime": "",
"enable_etimes": [
"23:59"
],
"enable_days_of_week": null,
"enable_days_of_weeks": [
[
"1",
"2",
"3",
@ -147,29 +320,65 @@
"5",
"6",
"0"
]
],
"enable_in_bg": 0,
"notify_recovered": 1,
"notify_channels": [
"email",
"dingtalk",
"wecom"
],
"notify_repeat_step": 60,
"notify_max_number": 0,
"recover_duration": 0,
"callbacks": [],
"runbook_url": "",
"append_tags": [],
"annotations": null
},
{
"cate": "prometheus",
"datasource_ids": [
0
],
"name": "packet loss in the inbound direction - categraf",
"note": "",
"prod": "metric",
"algorithm": "",
"algo_params": null,
"delay": 0,
"severity": 3,
"severities": [
3
],
"disabled": 0,
"prom_for_duration": 60,
"prom_ql": "increase(net_drop_in[1m]) > 0",
"rule_config": {
"algo_params": null,
"inhibit": false,
"prom_ql": "",
"queries": [
{
"prom_ql": "increase(net_drop_in[1m]) > 0",
"severity": 3
}
],
"notify_recovered": 1,
"notify_channels": [
"email",
"dingtalk",
"wecom"
],
"notify_repeat_step": 60,
"callbacks": [],
"runbook_url": "",
"append_tags": []
"severity": 0
},
{
"name": "packet loss in the inbound direction - categraf",
"note": "",
"severity": 3,
"disabled": 0,
"prom_for_duration": 60,
"prom_ql": "increase(net_drop_in[1m]) > 0",
"prom_eval_interval": 15,
"enable_stime": "00:00",
"enable_etime": "23:59",
"enable_days_of_week": [
"prom_eval_interval": 15,
"enable_stime": "",
"enable_stimes": [
"00:00"
],
"enable_etime": "",
"enable_etimes": [
"23:59"
],
"enable_days_of_week": null,
"enable_days_of_weeks": [
[
"1",
"2",
"3",
@ -177,29 +386,65 @@
"5",
"6",
"0"
]
],
"enable_in_bg": 0,
"notify_recovered": 1,
"notify_channels": [
"email",
"dingtalk",
"wecom"
],
"notify_repeat_step": 60,
"notify_max_number": 0,
"recover_duration": 0,
"callbacks": [],
"runbook_url": "",
"append_tags": [],
"annotations": null
},
{
"cate": "prometheus",
"datasource_ids": [
0
],
"name": "packet loss in the outbound direction - categraf",
"note": "",
"prod": "metric",
"algorithm": "",
"algo_params": null,
"delay": 0,
"severity": 3,
"severities": [
3
],
"disabled": 0,
"prom_for_duration": 60,
"prom_ql": "increase(net_drop_out[1m]) > 0",
"rule_config": {
"algo_params": null,
"inhibit": false,
"prom_ql": "",
"queries": [
{
"prom_ql": "increase(net_drop_out[1m]) > 0",
"severity": 3
}
],
"notify_recovered": 1,
"notify_channels": [
"email",
"dingtalk",
"wecom"
],
"notify_repeat_step": 60,
"callbacks": [],
"runbook_url": "",
"append_tags": []
"severity": 0
},
{
"name": "packet loss in the outbound direction - categraf",
"note": "",
"severity": 3,
"disabled": 0,
"prom_for_duration": 60,
"prom_ql": "increase(net_drop_out[1m]) > 0",
"prom_eval_interval": 15,
"enable_stime": "00:00",
"enable_etime": "23:59",
"enable_days_of_week": [
"prom_eval_interval": 15,
"enable_stime": "",
"enable_stimes": [
"00:00"
],
"enable_etime": "",
"enable_etimes": [
"23:59"
],
"enable_days_of_week": null,
"enable_days_of_weeks": [
[
"1",
"2",
"3",
@ -207,47 +452,159 @@
"5",
"6",
"0"
],
"notify_recovered": 1,
"notify_channels": [
"email",
"dingtalk",
"wecom"
],
"notify_repeat_step": 60,
"callbacks": [],
"runbook_url": "",
"append_tags": []
]
],
"enable_in_bg": 0,
"notify_recovered": 1,
"notify_channels": [
"email",
"dingtalk",
"wecom"
],
"notify_repeat_step": 60,
"notify_max_number": 0,
"recover_duration": 0,
"callbacks": [],
"runbook_url": "",
"append_tags": [],
"annotations": null
},
{
"cate": "prometheus",
"datasource_ids": [
0
],
"name": "大于200G的盘，空间不足了",
"note": "",
"prod": "metric",
"algorithm": "",
"algo_params": null,
"delay": 0,
"severity": 0,
"severities": [
3,
2,
1
],
"disabled": 0,
"prom_for_duration": 60,
"prom_ql": "",
"rule_config": {
"inhibit": true,
"queries": [
{
"prom_ql": "disk_free/1024/1024/1024 < 20 and disk_total/1024/1024/1024 >= 200",
"severity": 3
},
{
"prom_ql": "disk_free/1024/1024/1024 < 10 and disk_total/1024/1024/1024 >= 200",
"severity": 2
},
{
"prom_ql": "disk_free/1024/1024/1024 < 2 and disk_total/1024/1024/1024 >= 200",
"severity": 1
}
]
},
{
"name": "number of TME_WAIT exceeds 20,000 - categraf",
"note": "",
"severity": 2,
"disabled": 0,
"prom_for_duration": 60,
"prom_ql": "netstat_tcp_time_wait > 20000",
"prom_eval_interval": 15,
"enable_stime": "00:00",
"enable_etime": "23:59",
"enable_days_of_week": [
"prom_eval_interval": 30,
"enable_stime": "",
"enable_stimes": [
"00:00"
],
"enable_etime": "",
"enable_etimes": [
"23:59"
],
"enable_days_of_week": null,
"enable_days_of_weeks": [
[
"0",
"1",
"2",
"3",
"4",
"5",
"6",
"0"
],
"notify_recovered": 1,
"notify_channels": [
"email",
"dingtalk",
"wecom"
],
"notify_repeat_step": 60,
"callbacks": [],
"runbook_url": "",
"append_tags": []
}
]
"6"
]
],
"enable_in_bg": 0,
"notify_recovered": 1,
"notify_channels": [],
"notify_repeat_step": 60,
"notify_max_number": 0,
"recover_duration": 0,
"callbacks": [],
"runbook_url": "",
"append_tags": [],
"annotations": {}
},
{
"cate": "prometheus",
"datasource_ids": [
0
],
"name": "小于200G的盘，空间不足了",
"note": "",
"prod": "metric",
"algorithm": "",
"algo_params": null,
"delay": 0,
"severity": 0,
"severities": [
3,
2,
1
],
"disabled": 0,
"prom_for_duration": 60,
"prom_ql": "",
"rule_config": {
"inhibit": true,
"queries": [
{
"prom_ql": "disk_used_percent > 90 and disk_total/1024/1024/1024 < 200",
"severity": 3
},
{
"prom_ql": "disk_used_percent > 95 and disk_total/1024/1024/1024 < 200",
"severity": 2
},
{
"prom_ql": "disk_used_percent > 99 and disk_total/1024/1024/1024 < 200",
"severity": 1
}
]
},
"prom_eval_interval": 30,
"enable_stime": "",
"enable_stimes": [
"00:00"
],
"enable_etime": "",
"enable_etimes": [
"23:59"
],
"enable_days_of_week": null,
"enable_days_of_weeks": [
[
"0",
"1",
"2",
"3",
"4",
"5",
"6"
]
],
"enable_in_bg": 0,
"notify_recovered": 1,
"notify_channels": [],
"notify_repeat_step": 60,
"notify_max_number": 0,
"recover_duration": 0,
"callbacks": [],
"runbook_url": "",
"append_tags": [],
"annotations": {}
}
]