monitor_ob.yaml

obInputBasic: &obInputBasic
  plugin: mysqlTableInput
  config:
    timeout: 10s
    pluginConfig:
      maintainCacheThreads: 4
      connection:
        url: ${monagent.ob.monitor.user}:${monagent.ob.monitor.password}@tcp(127.0.0.1:${monagent.ob.sql.port})/oceanbase?interpolateParams=true
        maxIdle: 2
        maxOpen: 32
      defaultConditionValues:
        ob_svr_ip: ${monagent.host.ip}
        ob_svr_port: ${monagent.ob.rpc.port}
        ob_is_rootservice: false
        ob_is_v4: false
        ob_is_v3: false
      collectConfig:

        - sql: select ob_version() REGEXP '^4' as ob_is_v4
          name: ob_is_v4
          conditionValues:
            ob_is_v4: ob_is_v4
        - sql: select ob_version() REGEXP '^3' as ob_is_v3
          name: ob_is_v3
          conditionValues:
            ob_is_v3: ob_is_v3
        - name: ob_role
          sqlSlowThreshold: 100ms
          metrics:
            is_rootserver: with_rootserver
          conditionValues:
            ob_is_rootservice: with_rootserver
          conditions: [ ob_is_v3 ]
          params: [ ob_svr_ip, ob_svr_port ]
          sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) */ with_rootserver from __all_server where svr_ip = ? and svr_port = ?

        - name: ob_role
          sqlSlowThreshold: 100ms
          metrics:
            is_rootserver: with_rootserver
          conditionValues:
            ob_is_rootservice: with_rootserver
          conditions: [ ob_is_v4 ]
          params: [ ob_svr_ip, ob_svr_port ]
          sql: select (case when with_rootserver='YES' then 1 else 0 end) as with_rootserver from DBA_OB_SERVERS where svr_ip = ? and svr_port = ?

        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) */ tenant_name, cache_name, cache_size from __all_virtual_kvcache_info, __all_tenant where __all_tenant.tenant_id = __all_virtual_kvcache_info.tenant_id and svr_ip = ? and svr_port = ?
          params: [ob_svr_ip, ob_svr_port]
          name: ob_cache
          tags:
            tenant_name: tenant_name
            cache_name: cache_name
          metrics:
            size_bytes: cache_size
        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) */ tenant_name, tenant_id, stat_id, value from v$sysstat, __all_tenant where stat_id IN (10000, 10001, 10002, 10003, 10005, 10006, 140002, 140003, 140005, 140006, 60019, 60020, 60024, 80040, 80041, 130000, 130001, 130002, 130004, 20001, 20002, 30000, 30001, 30002, 30005, 30006, 40000, 40001, 40002, 40003, 40004, 40005, 40006, 40007, 40008, 40009, 40010, 40011, 40012, 50000, 50001, 50008, 50009, 60000, 60001, 60002, 60003, 60004, 60005, 60019, 60020, 60021, 60022, 60023, 60024, 80057) and (con_id > 1000 or con_id = 1) and __all_tenant.tenant_id = v$sysstat.con_id and class < 1000
          name: ob_sysstat
          tags:
            tenant_name: tenant_name
            tenant_id: tenant_id
            stat_id: stat_id
          metrics:
            value: value
        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) QUERY_TIMEOUT(100000000) */ tenant_name, 1 as role, case when cnt is null then 0 else cnt end as cnt from (select __all_tenant.tenant_name, t1.cnt from __all_tenant left join (select tenant_id, count(*) as cnt from __all_virtual_partition_info where svr_ip = ? and svr_port = ? group by tenant_id) t1 on __all_tenant.tenant_id = t1.tenant_id) t2
          params: [ob_svr_ip, ob_svr_port]
          conditions: [ ob_is_v3 ]
          name: ob_partition
          tags:
            tenant_name: tenant_name
            role: role
          metrics:
            num: cnt
          enableCache: true
          cacheExpire: 1h
          cacheDataExpire: 2h
        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) */ group_concat(svr_ip SEPARATOR ',') as servers, status, count(1) as cnt from __all_server group by status
          name: ob_server
          tags:
            server_ips: servers
            status: status
          metrics:
            num: cnt
          enableCache: true
          cacheExpire: 60s
          cacheDataExpire: 120s
        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) QUERY_TIMEOUT(100000000)  */ tenant_name, count(*)  as cnt from gv$table group by tenant_id
          name: ob_table
          conditions: [ ob_is_v3, ob_is_rootservice ]
          tags:
            tenant_name: tenant_name
          metrics:
            num: cnt
          enableCache: true
          cacheExpire: 1h
          cacheDataExpire: 2h
        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) QUERY_TIMEOUT(100000000)  */ tenant_name, count(*)  as cnt from CDB_TABLES join __all_tenant on CDB_TABLES.con_id = __all_tenant.tenant_id group by con_id
          name: ob_table
          conditions: [ ob_is_v4, ob_is_rootservice ]
          tags:
            tenant_name: tenant_name
          metrics:
            num: cnt
          enableCache: true
          cacheExpire: 1h
          cacheDataExpire: 2h
        - sql: select  /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) */ case when cnt is null then 0 else cnt end as cnt, tenant_name from (select __all_tenant.tenant_name, cnt from __all_tenant left join (select count(*) as cnt, tenant as tenant_name from __all_virtual_processlist where svr_ip = ? and svr_port = ? group by tenant) t1 on __all_tenant.tenant_name = t1.tenant_name) t2
          params: [ob_svr_ip, ob_svr_port]
          name: ob_active_session
          tags:
            tenant_name: tenant_name
          metrics:
            num: cnt
        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) */ zone, name, value, time_to_usec(now()) as current from __all_zone
          name: ob_zone
          tags:
            zone: zone
            name: name
          metrics:
            stat: value
            current_timestamp: current
        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) */ tenant_name, mem_used, access_count, hit_count from v$plan_cache_stat join __all_tenant on v$plan_cache_stat.tenant_id = __all_tenant.tenant_id group by tenant_name
          name: ob_plan_cache
          conditions: [ ob_is_v3 ]
          tags:
            tenant_name: tenant_name
          metrics:
            memory_bytes: mem_used
            access_total: access_count
            hit_total: hit_count
        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) */ tenant_name, mem_used, access_count, hit_count from V$OB_PLAN_CACHE_STAT join __all_tenant on V$OB_PLAN_CACHE_STAT.tenant_id = __all_tenant.tenant_id group by tenant_name
          name: ob_plan_cache
          condition: [ ob_is_v4 ]

          tags:
            tenant_name: tenant_name
          metrics:
            memory_bytes: mem_used
            access_total: access_count
            hit_total: hit_count
        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) */ tenant_name, sum(total_waits) as total_waits, sum(time_waited_micro) / 1000000 as time_waited from v$system_event join __all_tenant on v$system_event.con_id = __all_tenant.tenant_id where v$system_event.wait_class <> 'IDLE' group by tenant_name
          name: ob_waitevent
          tags:
            tenant_name: tenant_name
          metrics:
            wait_total: total_waits
            wait_seconds_total: time_waited
        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) */ __all_tenant.tenant_name as tenant_name, cast(v_acc_response_time.response_time / 1000000 as float) as response_time_seconds,  v_acc_response_time.count as bucket, case when  v_acc_response_time.response_time = (select max(response_time) from __all_virtual_query_response_time) then v_acc_response_time.count else null end as count, case when  v_acc_response_time.response_time = (select max(response_time) from __all_virtual_query_response_time) then cast(v_acc_response_time.sum / 1000000 as float) else null end as sum from (select b.tenant_id, b.response_time  as response_time, sum(a.count) as count, sum(a.total) as sum from  __all_virtual_query_response_time a,  __all_virtual_query_response_time b where a.response_time <= b.response_time and a.svr_ip = b.svr_ip and a.svr_port = b.svr_port and b.svr_ip = ? and b.svr_port = ? group by b.tenant_id, b.response_time) v_acc_response_time,  __all_tenant where v_acc_response_time.tenant_id = __all_tenant.tenant_id;
          params: [ob_svr_ip, ob_svr_port]
          name: ob_query_response_time_seconds
          tags:
            tenant_name: tenant_name
            le: response_time_seconds
          metrics:
            bucket: bucket
            count: count
            sum: sum

obInputExtra: &obInputExtra
  plugin: mysqlTableInput
  config:
    timeout: 10s
    pluginConfig:
      maintainCacheThreads: 4
      connection:
        url: ${monagent.ob.monitor.user}:${monagent.ob.monitor.password}@tcp(127.0.0.1:${monagent.ob.sql.port})/oceanbase?interpolateParams=true
        maxIdle: 2
        maxOpen: 32
      defaultConditionValues:
        ob_svr_ip: ${monagent.host.ip}
        ob_svr_port: ${monagent.ob.rpc.port}
        ob_is_rootservice: false
        ob_is_v4: false
        ob_is_v3: false
      collectConfig:
        - sql: select ob_version() REGEXP '^4' as ob_is_v4
          name: ob_is_v4
          conditionValues:
            ob_is_v4: ob_is_v4
        - sql: select ob_version() REGEXP '^3' as ob_is_v3
          name: ob_is_v3
          conditionValues:
            ob_is_v3: ob_is_v3
        - name: ob_role
          sqlSlowThreshold: 100ms
          metrics:
            is_rootserver: with_rootserver
          conditionValues:
            ob_is_rootservice: with_rootserver
          conditions: [ ob_is_v3 ]
          params: [ ob_svr_ip, ob_svr_port ]
          sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) */ with_rootserver from __all_server where svr_ip = ? and svr_port = ?

        - name: ob_role
          sqlSlowThreshold: 100ms
          metrics:
            is_rootserver: with_rootserver
          conditionValues:
            ob_is_rootservice: with_rootserver
          conditions: [ ob_is_v4 ]
          params: [ ob_svr_ip, ob_svr_port ]
          sql: select (case when with_rootserver='YES' then 1 else 0 end) as with_rootserver from DBA_OB_SERVERS where svr_ip = ? and svr_port = ?
        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) */ count(*) as cnt from v$unit
          name: ob_unit
          conditions: [ ob_is_v3 ]
          metrics:
            num: cnt
        - sql: select /*+ MONITOR_AGENT */ count(*) as cnt from V$OB_UNITS
          name: ob_unit
          conditions: [ ob_is_v4 ]
          metrics:
            num: cnt
        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) */ total_size, free_size from __all_virtual_disk_stat where svr_ip = ? and svr_port = ?
          params: [ob_svr_ip, ob_svr_port]
          name: ob_disk
          metrics:
            total_bytes: total_size
            free_bytes: free_size
          enableCache: true
          cacheExpire: 1h
          cacheDataExpire: 2h
        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) */ cpu_total,cpu_assigned,mem_total,mem_assigned,disk_total,disk_assigned,unit_num,migrating_unit_num,cpu_assigned_percent, mem_assigned_percent, disk_assigned_percent from __all_virtual_server_stat where svr_ip = ? and svr_port = ?
          params: [ob_svr_ip, ob_svr_port]
          name: ob_server_resource
          conditions: [ ob_is_v3 ]
          metrics:
            cpu: cpu_total
            cpu_assigned: cpu_assigned
            memory_bytes: mem_total
            memory_assigned_bytes: mem_assigned
            disk_bytes: disk_total
            disk_assigned_bytes: disk_assigned
            unit_num: unit_num
            migrating_unit_num: migrating_unit_num
            cpu_assigned_percent: cpu_assigned_percent
            memory_assigned_percent: mem_assigned_percent
            disk_assigned_percent: disk_assigned_percent
          enableCache: true
          cacheExpire: 60s
          cacheDataExpire: 120s
        - sql: select /* MONITOR_AGENT */ cpu_capacity_max as cpu_total,cpu_assigned_max as cpu_assigned,mem_capacity as mem_total,mem_assigned as mem_assigned,data_disk_capacity as disk_total, (cpu_assigned_max / cpu_capacity_max) as cpu_assigned_percent, (mem_assigned / mem_capacity) as mem_assigned_percent from GV$OB_SERVERS where svr_ip = ? and svr_port = ?
          params: [ob_svr_ip, ob_svr_port]
          name: ob_server_resource
          conditions: [ ob_is_v4 ]
          metrics:
            cpu: cpu_total
            cpu_assigned: cpu_assigned
            memory_bytes: mem_total
            memory_assigned_bytes: mem_assigned
            disk_bytes: disk_total
            cpu_assigned_percent: cpu_assigned_percent
            memory_assigned_percent: mem_assigned_percent
          enableCache: true
          cacheExpire: 60s
          cacheDataExpire: 120s
        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) */ count(1) as cnt from __all_virtual_trans_stat where part_trans_action > 2 and ctx_create_time < date_sub(now(), interval 600 second) and svr_ip = ? and svr_port = ?
          params: [ob_svr_ip, ob_svr_port]
          name: ob_trans
          conditions: [ ob_is_rootservice ]
          metrics:
            expire_num: cnt
        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) QUERY_TIMEOUT(100000000)  */ count(*) as cnt from gv$table where table_type in (5) and index_status in (5, 6)
          name: ob_index
          conditions: [ ob_is_v3, ob_is_rootservice]
          metrics:
            error_num: cnt
        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) QUERY_TIMEOUT(100000000)  */ count(*) as cnt from CDB_INDEXES where table_type in (5) and status in (5, 6)
          name: ob_index
          conditions: [ob_is_v4, ob_is_rootservice ]
          metrics:
            error_num: cnt
        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) */ __all_tenant.tenant_name,  active, total, freeze_trigger, freeze_cnt from gv$memstore, __all_tenant  where gv$memstore.tenant_id = __all_tenant.tenant_id and ip = ? and port = ?
          params: [ob_svr_ip, ob_svr_port]
          name: ob_memstore
          conditions: [ ob_is_v3 ]
          tags:
            tenant_name: tenant_name
          metrics:
            active_bytes: active
            total_bytes: total
            freeze_trigger_bytes: freeze_trigger
            freeze_times: freeze_cnt
        - sql: select /* MONITOR_AGENT */ __all_tenant.tenant_name, active_span as active, memstore_used as total, freeze_trigger, freeze_cnt from GV$OB_MEMSTORE, __all_tenant where GV$OB_MEMSTORE.tenant_id = __all_tenant.tenant_id and svr_ip = ? and svr_port = ?
          params: [ob_svr_ip, ob_svr_port]
          name: ob_memstore
          conditions: [ ob_is_v4 ]
          tags:
            tenant_name: tenant_name
          metrics:
            active_bytes: active
            total_bytes: total
            freeze_trigger_bytes: freeze_trigger
            freeze_times: freeze_cnt
        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) */ sum(hold) as hold, sum(used) as used from __all_virtual_memory_info where tenant_id = 500 and svr_ip = ? and svr_port = ?
          params: [ob_svr_ip, ob_svr_port]
          name: ob_tenant500_memory
          metrics:
            hold_bytes: hold
            used_bytes: used
        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) */ tenant_name, __all_unit_config.name, max_cpu, min_cpu, max_memory, min_memory, max_iops, min_iops, max_disk_size, max_session_num from __all_tenant, __all_resource_pool, __all_unit_config, __all_unit  where __all_tenant.tenant_id = __all_resource_pool.tenant_id and __all_resource_pool.unit_config_id = __all_unit_config.unit_config_id and __all_unit.resource_pool_id = __all_resource_pool.resource_pool_id and __all_unit.svr_ip = ? and __all_unit.svr_port = ?
          params: [ob_svr_ip, ob_svr_port]
          name: ob_unit_config
          conditions: [ ob_is_v3 ]
          tags:
            tenant_name: tenant_name
            unit_config_name: name
          metrics:
            max_cpu: max_cpu
            min_cpu: min_cpu
            max_memory_bytes: max_memory
            min_memory_bytes: min_memory
            max_iops: max_iops
            min_iops: min_iops
            max_disk_size_bytes: max_disk_size
            max_session_num: max_session_num
        - sql: select /* MONITOR_AGENT */ __all_tenant.tenant_name, DBA_OB_UNIT_CONFIGS.name, DBA_OB_UNIT_CONFIGS.max_cpu, DBA_OB_UNIT_CONFIGS.min_cpu, DBA_OB_UNIT_CONFIGS.memory_size as max_memory, DBA_OB_UNIT_CONFIGS.max_iops, DBA_OB_UNIT_CONFIGS.min_iops from __all_tenant, DBA_OB_RESOURCE_POOLS, DBA_OB_UNIT_CONFIGS, DBA_OB_UNITS  where DBA_OB_RESOURCE_POOLS.tenant_id = __all_tenant.tenant_id and DBA_OB_RESOURCE_POOLS.unit_config_id = DBA_OB_UNIT_CONFIGS.unit_config_id and DBA_OB_UNITS.resource_pool_id = DBA_OB_RESOURCE_POOLS.resource_pool_id and DBA_OB_UNITS.svr_ip = ? and DBA_OB_UNITS.svr_port = ?
          params: [ob_svr_ip, ob_svr_port]
          name: ob_unit_config
          conditions: [ ob_is_v4 ]
          tags:
            tenant_name: tenant_name
            unit_config_name: name
          metrics:
            max_cpu: max_cpu
            min_cpu: min_cpu
            max_memory_bytes: max_memory
            min_memory_bytes: min_memory
            max_iops: max_iops
            min_iops: min_iops
          enableCache: true
          cacheExpire: 60s
        - sql: select /*+ MONITOR_AGENT READ_CONSISTENCY(WEAK) */ tenant_name, tenant_id, `latch#` as latch_no, name, gets, misses, sleeps, immediate_gets, immediate_misses, spin_gets, wait_time / 1000000 as wait_time from v$latch, __all_tenant where __all_tenant.tenant_id = v$latch.con_id
          name: ob_latch
          tags:
            tenant_name: tenant_name
            tenant_id: tenant_id
            name: name
            latch_no: latch_no
          metrics:
            get_total: gets
            miss_total: misses
            sleep_total: sleeps
            immediate_get_total: immediate_gets
            immediate_miss_total: immediate_misses
            spin_get_total: spin_gets
            wait_seconds_total: wait_time

retagProcessor: &retagProcessor
  plugin: retagProcessor
  config:
    timout: 10s
    pluginConfig:
      newTags:
        app: OB
        obzone: ${monagent.ob.zone.name}
        svr_ip: ${monagent.host.ip}
        ob_cluster_id: ${monagent.ob.cluster.id}
        ob_cluster_name: ${monagent.ob.cluster.name}

prometheusExporter: &prometheusExporter
  plugin: prometheusExporter
  config:
    timout: 10s
    pluginConfig:
      formatType: fmtText

modules:
  - module: monitor.ob
    moduleType: monagent.pipeline
    process: monagent
    config:
      name: monitor.ob
      status: ${monagent.pipeline.ob.status}
      pipelines:
        - name: ob_basic
          config:
            scheduleStrategy: trigger
            exposeUrl: /metrics/ob/basic
          structure:
            inputs:
              - <<: *obInputBasic
            processors:
              - <<: *retagProcessor
            exporter:
              <<: *prometheusExporter
        - name: ob_extra
          config:
            scheduleStrategy: trigger
            exposeUrl: /metrics/ob/extra
          structure:
            inputs:
              - <<: *obInputExtra
            processors:
              - <<: *retagProcessor
            exporter:
              <<: *prometheusExporter