人大金仓学习之四-kmonitor

济南小老虎 / 2023-07-21 / 原文

人大金仓学习之四-kmonitor


背景

kmonitor 其实时一个很好的工具和思路
开元的软件封装一下, 减轻技术复杂度,提高部署效率
并且能够更加快速的定位解决问题. 能够极大的提升客户体验.

并且界面比较好看, 比kwr 看起来高大上多了. 
不仅可以解决部分实际问题, 也可以提高产品的产品力

下载地址

需要注意 这个组件官方没有下载地址,
需要联系原厂, 获取对应的安装截止. 

截止里面其实包含一套 grafana的安装和默认配置
包含一套java写的 kingbase_exporter
以及默认安装的一套prometheus
外加一个node_exporter

需要注意, 这个安装介质时区分架构和平台的. 

部署文档

2.1 创建kmonitor用户
使⽤system⽤户登陆监控⽬标数据库,执⾏scripts/KmonitorUserInit.sql中语句,创建kmonitor⽤户并进⾏授权
-- 创建监控用户
CREATE USER kmonitor with password 'kmonitor';
-- 添加用户标注
COMMENT ON ROLE kmonitor IS 'kingbase KMonitor user';
-- 授权监控用户
GRANT pg_monitor TO kmonitor;  -- V008R006 
GRANT SYS_SIGNAL_BACKEND TO kmonitor;  -- V008R003 
-- 授权监控用户使用kwr扩展
GRANT USAGE ON SCHEMA perf TO kmonitor;

如未开启kwr插件或不支持kwr插件版本执行以下语句创建perf schema:
CREATE SCHEMA IF NOT EXISTS perf;


Oracle模式下创建tps和qps的SQL

注:oracle模式执行
-- V008R006(oracle mode)

create or replace function perf.tps()
 returns int as $$
declare
 v1 int8;
 v2 int8;
 v3 int8;
begin
 select sum(xact_commit + xact_rollback) from sys_stat_database into v1;
 perform sys_sleep(1);
 perform sys_stat_clear_snapshot();
 select sum(xact_commit + xact_rollback) from sys_stat_database into v2;
 select v2-v1 into v3;
 return v3;
end;
$$ language plpgsql;

create or replace function perf.qps()
 returns int as $$
declare
 v1 int8;
 v2 int8;
 v3 int8;
begin
 select sum(calls) from sys_stat_msgaccum
 where message = 'Simple Query' or message = 'Execute' into v1;
 perform sys_sleep(1);
 perform sys_stat_clear_snapshot();
 select sum(calls) from sys_stat_msgaccum
 where message = 'Simple Query' or message = 'Execute' into v2;
 select v2-v1 into v3;
 return v3;
end;
$$ language plpgsql;


修改配置文件

shared_preload_libraries = 'liboracle_parser, synonym, plsql, force_view, kdb_flashback,plugin_debugger, plsql_plugin_debugger, plsql_plprofiler, ora_commands,kdb_ora_expr, sepapower, dblink, sys_kwr, sys_spacequota, sys_stat_statements, backtrace, kdb_utils_function, auto_bmr, sys_squeeze, src_restrict'
sys_stat_statements.track='all'

# 这里面再添加几个kwr需要使用的参数, 可以一并使用
track_activities = on
track_sql = on
track_instance = on
track_wait_timing = on
track_counts = on
track_io_timing = on
track_functions = 'all'
sys_stat_statements.track = 'top'
sys_stat_statements.track_parse = on
sys_stat_statements.track_plan = on
sys_kwr.enable = on
sys_kwr.history_days = 14
sys_kwr.interval = 30

可以用来生成 kwr报告:
select * from perf.create_snapshot();

再次执行
select * from perf.create_snapshot();

select * from perf.kwr_report_to_file(2,3,'html','/nvme01/kingbase/kwr/zhaobsh_002_003.html');

这个SQL可以查看表大小信息

select schemaname,relname,sys_size_pretty(sys_total_relation_size(relid)) from sys_stat_user_tables order by sys_relation_size(relid) desc limit 30;

---
## kmonitor的相关设置

其实只需要修改一个文件就可以了, 但是可能需要执行两个文件.
需要注意 kmonitor不一定需要运行到kingbase上面
可以在其他机器上面运行. 但是链接人大金仓就可以.

kmonitor server ("kmonitor" "prometheus" "consul" "kingbase_exporter" "node_exporter" "alertmanager")

KMONITOR_SERVER=("kmonitor" "prometheus" "consul" "kingbase_exporter" "node_exporter")

kmonitor config

KMONITOR_PORT=3000

prometheus config

PROMETHEUS_PORT=9090
STORAGE_TSDB_PATH="data"
STORAGE_TSDB_RETENTION_TIME="15d"
STORAGE_TSDB_RETENTION_SIZE="512MB"

kingbase_exporter config

KINGBASE_EXPORTER_PORT=1234

node_exporter config

NODE_EXPORTER_PORT=9100

consul config

CONSUL_SERVER_PORT=8300
CONSUL_SERF_LAN_PORT=8301
CONSUL_SERF_WAN_PORT=8202
CONSUL_HTTP_PORT=8500
CONSUL_DNS_PORT=8600
CONSUL_DATA="consul_data"

kingbase config

KINGBASE_ADDRESS="192.168.255.xxx"
KINGBASE_PORT=54321
KINGBASE_DB="test"
KINGBASE_USER="kmonitor"
KINGBASE_PASSWD="xxxx"

alertmanager config

ALERTMANAGER_PORT=9093
ALERTMANAGER_CLUSTER_PORT=9094

local install enable(1) disable(0)

KINGBASE_EXPORTER_LOCAL_INSTALL=1
NODE_EXPORTER_LOCAL_INSTALL=1
ALERTMANAGER_LOCAL_INSTALL=0

excluded application

EXCLUDED_PROCESS=("")

register config

CLUSTER="Kingbase集群"
USER="kingbase"
REGISTER_KINGBASE_LIST=("192.168.255.yyy_1234")
REGISTER_NODE_LIST=("192.168.255.xxx_9100")

deregister config

DEREGISTER_LIST=()