监控系统

infra:            # 代理、监控、警报等的infra集群..
  hosts: { 10.10.10.10: { infra_seq: 1 } }
  vars:
    pg_exporters:   # 在此列出所有待监控的远程 RDS PG 实例

      20001:        # 分配一个唯一的未使用的本地端口，供本地监控 Agent 使用，这里是一个 PolarDB 的主库
        pg_cluster: pg-polar                  # RDS 集群名 （身份参数，手工指定分配监控系统内名称）
        pg_seq: 1                             # RDS 实例号 （身份参数，手工指定分配监控系统内名称）
        pg_host: pc-2ze379wb1d4irc18x.polardbpg.rds.aliyuncs.com # RDS 主机地址
        pg_port: 1921                         # RDS 端口（从控制台连接信息获取）
        pg_exporter_auto_discovery: true      # 启用新数据库自动发现功能
        pg_exporter_include_database: 'test'  # 仅监控这个列表中的数据库（多个数据库用逗号分隔）
        pg_monitor_username: dbuser_monitor   # 监控用的用户名，覆盖全局配置
        pg_monitor_password: DBUser_Monitor   # 监控用的密码，覆盖全局配置
        pg_databases: [{ name: test }]        # 希望启用PGCAT的数据库列表，只要name字段即可，register_datasource设置为false则不注册。

      20002:       # 这是一个 PolarDB  从库
        pg_cluster: pg-polar                  # RDS 集群名 （身份参数，手工指定分配监控系统内名称）
        pg_seq: 2                             # RDS 实例号 （身份参数，手工指定分配监控系统内名称）
        pg_host: pe-2ze7tg620e317ufj4.polarpgmxs.rds.aliyuncs.com # RDS 主机地址
        pg_port: 1521                         # RDS 端口（从控制台连接信息获取）
        pg_exporter_auto_discovery: true      # 启用新数据库自动发现功能
        pg_exporter_include_database: 'test,postgres'  # 仅监控这个列表中的数据库（多个数据库用逗号分隔）
        pg_monitor_username: dbuser_monitor   # 监控用的用户名
        pg_monitor_password: DBUser_Monitor   # 监控用的密码
        pg_databases: [ { name: test } ]        # 希望启用PGCAT的数据库列表，只要name字段即可，register_datasource设置为false则不注册。

      20004: # 这是一个基础版的单节点 RDS for PostgreSQL 实例
        pg_cluster: pg-rds                    # RDS 集群名 （身份参数，手工指定分配监控系统内名称）
        pg_seq: 1                             # RDS 实例号 （身份参数，手工指定分配监控系统内名称）
        pg_host: pgm-2zern3d323fe9ewk.pg.rds.aliyuncs.com  # RDS 主机地址
        pg_port: 5432                         # RDS 端口（从控制台连接信息获取）
        pg_exporter_auto_discovery: true      # 启用新数据库自动发现功能
        pg_exporter_include_database: 'rds'   # 仅监控这个列表中的数据库（多个数据库用逗号分隔）
        pg_monitor_username: dbuser_monitor   # 监控用的用户名
        pg_monitor_password: DBUser_Monitor   # 监控用的密码
        pg_databases: [ { name: rds } ]       # 希望启用PGCAT的数据库列表，只要name字段即可，register_datasource设置为false则不注册。

      20005: # 这是一个高可用版的 RDS for PostgreSQL 集群主库
        pg_cluster: pg-rdsha                  # RDS 集群名 （身份参数，手工指定分配监控系统内名称）
        pg_seq: 1                             # RDS 实例号 （身份参数，手工指定分配监控系统内名称）
        pg_host: pgm-2ze3d35d27bq08wu.pg.rds.aliyuncs.com  # RDS 主机地址
        pg_port: 5432                         # RDS 端口（从控制台连接信息获取）
        pg_exporter_include_database: 'rds'   # 仅监控这个列表中的数据库（多个数据库用逗号分隔）
        pg_databases: [ { name: rds }, {name : test} ]  # 将这两个数据库纳入 PGCAT 管理，注册为 Grafana 数据源

      20006: # 这是一个高可用版的 RDS for PostgreSQL 集群只读实例（从库）
        pg_cluster: pg-rdsha                  # RDS 集群名 （身份参数，手工指定分配监控系统内名称）
        pg_seq: 2                             # RDS 实例号 （身份参数，手工指定分配监控系统内名称）
        pg_host: pgr-2zexqxalk7d37edt.pg.rds.aliyuncs.com  # RDS 主机地址
        pg_port: 5432                         # RDS 端口（从控制台连接信息获取）
        pg_exporter_include_database: 'rds'   # 仅监控这个列表中的数据库（多个数据库用逗号分隔）
        pg_databases: [ { name: rds }, {name : test} ]  # 将这两个数据库纳入 PGCAT 管理，注册为 Grafana 数据源

----------------------------------------------------------------------
-- Table bloat estimate : monitor.pg_table_bloat
----------------------------------------------------------------------
DROP VIEW IF EXISTS monitor.pg_table_bloat CASCADE;
CREATE OR REPLACE VIEW monitor.pg_table_bloat AS
SELECT CURRENT_CATALOG AS datname, nspname, relname , tblid , bs * tblpages AS size,
       CASE WHEN tblpages - est_tblpages_ff > 0 THEN (tblpages - est_tblpages_ff)/tblpages::FLOAT ELSE 0 END AS ratio
FROM (
         SELECT ceil( reltuples / ( (bs-page_hdr)*fillfactor/(tpl_size*100) ) ) + ceil( toasttuples / 4 ) AS est_tblpages_ff,
                tblpages, fillfactor, bs, tblid, nspname, relname, is_na
         FROM (
                  SELECT
                      ( 4 + tpl_hdr_size + tpl_data_size + (2 * ma)
                          - CASE WHEN tpl_hdr_size % ma = 0 THEN ma ELSE tpl_hdr_size % ma END
                          - CASE WHEN ceil(tpl_data_size)::INT % ma = 0 THEN ma ELSE ceil(tpl_data_size)::INT % ma END
                          ) AS tpl_size, (heappages + toastpages) AS tblpages, heappages,
                      toastpages, reltuples, toasttuples, bs, page_hdr, tblid, nspname, relname, fillfactor, is_na
                  FROM (
                           SELECT
                               tbl.oid AS tblid, ns.nspname , tbl.relname, tbl.reltuples,
                               tbl.relpages AS heappages, coalesce(toast.relpages, 0) AS toastpages,
                               coalesce(toast.reltuples, 0) AS toasttuples,
                               coalesce(substring(array_to_string(tbl.reloptions, ' ') FROM 'fillfactor=([0-9]+)')::smallint, 100) AS fillfactor,
                               current_setting('block_size')::numeric AS bs,
                               CASE WHEN version()~'mingw32' OR version()~'64-bit|x86_64|ppc64|ia64|amd64' THEN 8 ELSE 4 END AS ma,
                               24 AS page_hdr,
                               23 + CASE WHEN MAX(coalesce(s.null_frac,0)) > 0 THEN ( 7 + count(s.attname) ) / 8 ELSE 0::int END
                                   + CASE WHEN bool_or(att.attname = 'oid' and att.attnum < 0) THEN 4 ELSE 0 END AS tpl_hdr_size,
                               sum( (1-coalesce(s.null_frac, 0)) * coalesce(s.avg_width, 0) ) AS tpl_data_size,
                               bool_or(att.atttypid = 'pg_catalog.name'::regtype)
                                   OR sum(CASE WHEN att.attnum > 0 THEN 1 ELSE 0 END) <> count(s.attname) AS is_na
                           FROM pg_attribute AS att
                                    JOIN pg_class AS tbl ON att.attrelid = tbl.oid
                                    JOIN pg_namespace AS ns ON ns.oid = tbl.relnamespace
                                    LEFT JOIN pg_stats AS s ON s.schemaname=ns.nspname AND s.tablename = tbl.relname AND s.inherited=false AND s.attname=att.attname
                                    LEFT JOIN pg_class AS toast ON tbl.reltoastrelid = toast.oid
                           WHERE NOT att.attisdropped AND tbl.relkind = 'r' AND nspname NOT IN ('pg_catalog','information_schema')
                           GROUP BY 1,2,3,4,5,6,7,8,9,10
                       ) AS s
              ) AS s2
     ) AS s3
WHERE NOT is_na;
COMMENT ON VIEW monitor.pg_table_bloat IS 'postgres table bloat estimate';

GRANT SELECT ON monitor.pg_table_bloat TO pg_monitor;

----------------------------------------------------------------------
-- Index bloat estimate : monitor.pg_index_bloat
----------------------------------------------------------------------
DROP VIEW IF EXISTS monitor.pg_index_bloat CASCADE;
CREATE OR REPLACE VIEW monitor.pg_index_bloat AS
SELECT CURRENT_CATALOG AS datname, nspname, idxname AS relname, tblid, idxid, relpages::BIGINT * bs AS size,
       COALESCE((relpages - ( reltuples * (6 + ma - (CASE WHEN index_tuple_hdr % ma = 0 THEN ma ELSE index_tuple_hdr % ma END)
                                               + nulldatawidth + ma - (CASE WHEN nulldatawidth % ma = 0 THEN ma ELSE nulldatawidth % ma END))
                                  / (bs - pagehdr)::FLOAT  + 1 )), 0) / relpages::FLOAT AS ratio
FROM (
         SELECT nspname,idxname,indrelid AS tblid,indexrelid AS idxid,
                reltuples,relpages,
                current_setting('block_size')::INTEGER                                                               AS bs,
                (CASE WHEN version() ~ 'mingw32' OR version() ~ '64-bit|x86_64|ppc64|ia64|amd64' THEN 8 ELSE 4 END)  AS ma,
                24                                                                                                   AS pagehdr,
                (CASE WHEN max(COALESCE(pg_stats.null_frac, 0)) = 0 THEN 2 ELSE 6 END)                               AS index_tuple_hdr,
                sum((1.0 - COALESCE(pg_stats.null_frac, 0.0)) *
                    COALESCE(pg_stats.avg_width, 1024))::INTEGER                                                     AS nulldatawidth
         FROM pg_attribute
                  JOIN (
             SELECT pg_namespace.nspname,
                    ic.relname                                                   AS idxname,
                    ic.reltuples,
                    ic.relpages,
                    pg_index.indrelid,
                    pg_index.indexrelid,
                    tc.relname                                                   AS tablename,
                    regexp_split_to_table(pg_index.indkey::TEXT, ' ') :: INTEGER AS attnum,
                    pg_index.indexrelid                                          AS index_oid
             FROM pg_index
                      JOIN pg_class ic ON pg_index.indexrelid = ic.oid
                      JOIN pg_class tc ON pg_index.indrelid = tc.oid
                      JOIN pg_namespace ON pg_namespace.oid = ic.relnamespace
                      JOIN pg_am ON ic.relam = pg_am.oid
             WHERE pg_am.amname = 'btree' AND ic.relpages > 0 AND nspname NOT IN ('pg_catalog', 'information_schema')
         ) ind_atts ON pg_attribute.attrelid = ind_atts.indexrelid AND pg_attribute.attnum = ind_atts.attnum
                  JOIN pg_stats ON pg_stats.schemaname = ind_atts.nspname
             AND ((pg_stats.tablename = ind_atts.tablename AND pg_stats.attname = pg_get_indexdef(pg_attribute.attrelid, pg_attribute.attnum, TRUE))
                 OR (pg_stats.tablename = ind_atts.idxname AND pg_stats.attname = pg_attribute.attname))
         WHERE pg_attribute.attnum > 0
         GROUP BY 1, 2, 3, 4, 5, 6
     ) est;
COMMENT ON VIEW monitor.pg_index_bloat IS 'postgres index bloat estimate (btree-only)';

GRANT SELECT ON monitor.pg_index_bloat TO pg_monitor;

----------------------------------------------------------------------
-- Relation Bloat : monitor.pg_bloat
----------------------------------------------------------------------
DROP VIEW IF EXISTS monitor.pg_bloat CASCADE;
CREATE OR REPLACE VIEW monitor.pg_bloat AS
SELECT coalesce(ib.datname, tb.datname)                                                   AS datname,
       coalesce(ib.nspname, tb.nspname)                                                   AS nspname,
       coalesce(ib.tblid, tb.tblid)                                                       AS tblid,
       coalesce(tb.nspname || '.' || tb.relname, ib.nspname || '.' || ib.tblid::RegClass) AS tblname,
       tb.size                                                                            AS tbl_size,
       CASE WHEN tb.ratio < 0 THEN 0 ELSE round(tb.ratio::NUMERIC, 6) END                 AS tbl_ratio,
       (tb.size * (CASE WHEN tb.ratio < 0 THEN 0 ELSE tb.ratio::NUMERIC END)) ::BIGINT    AS tbl_wasted,
       ib.idxid,
       ib.nspname || '.' || ib.relname                                                    AS idxname,
       ib.size                                                                            AS idx_size,
       CASE WHEN ib.ratio < 0 THEN 0 ELSE round(ib.ratio::NUMERIC, 5) END                 AS idx_ratio,
       (ib.size * (CASE WHEN ib.ratio < 0 THEN 0 ELSE ib.ratio::NUMERIC END)) ::BIGINT    AS idx_wasted
FROM monitor.pg_index_bloat ib
         FULL OUTER JOIN monitor.pg_table_bloat tb ON ib.tblid = tb.tblid;

COMMENT ON VIEW monitor.pg_bloat IS 'postgres relation bloat detail';
GRANT SELECT ON monitor.pg_bloat TO pg_monitor;

----------------------------------------------------------------------
-- monitor.pg_index_bloat_human
----------------------------------------------------------------------
DROP VIEW IF EXISTS monitor.pg_index_bloat_human CASCADE;
CREATE OR REPLACE VIEW monitor.pg_index_bloat_human AS
SELECT idxname                            AS name,
       tblname,
       idx_wasted                         AS wasted,
       pg_size_pretty(idx_size)           AS idx_size,
       round(100 * idx_ratio::NUMERIC, 2) AS idx_ratio,
       pg_size_pretty(idx_wasted)         AS idx_wasted,
       pg_size_pretty(tbl_size)           AS tbl_size,
       round(100 * tbl_ratio::NUMERIC, 2) AS tbl_ratio,
       pg_size_pretty(tbl_wasted)         AS tbl_wasted
FROM monitor.pg_bloat
WHERE idxname IS NOT NULL;
COMMENT ON VIEW monitor.pg_index_bloat_human IS 'postgres index bloat info in human-readable format';
GRANT SELECT ON monitor.pg_index_bloat_human TO pg_monitor;


----------------------------------------------------------------------
-- monitor.pg_table_bloat_human
----------------------------------------------------------------------
DROP VIEW IF EXISTS monitor.pg_table_bloat_human CASCADE;
CREATE OR REPLACE VIEW monitor.pg_table_bloat_human AS
SELECT tblname                                          AS name,
       idx_wasted + tbl_wasted                          AS wasted,
       pg_size_pretty(idx_wasted + tbl_wasted)          AS all_wasted,
       pg_size_pretty(tbl_wasted)                       AS tbl_wasted,
       pg_size_pretty(tbl_size)                         AS tbl_size,
       tbl_ratio,
       pg_size_pretty(idx_wasted)                       AS idx_wasted,
       pg_size_pretty(idx_size)                         AS idx_size,
       round(idx_wasted::NUMERIC * 100.0 / idx_size, 2) AS idx_ratio
FROM (SELECT datname,
             nspname,
             tblname,
             coalesce(max(tbl_wasted), 0)                         AS tbl_wasted,
             coalesce(max(tbl_size), 1)                           AS tbl_size,
             round(100 * coalesce(max(tbl_ratio), 0)::NUMERIC, 2) AS tbl_ratio,
             coalesce(sum(idx_wasted), 0)                         AS idx_wasted,
             coalesce(sum(idx_size), 1)                           AS idx_size
      FROM monitor.pg_bloat
      WHERE tblname IS NOT NULL
      GROUP BY 1, 2, 3
     ) d;
COMMENT ON VIEW monitor.pg_table_bloat_human IS 'postgres table bloat info in human-readable format';
GRANT SELECT ON monitor.pg_table_bloat_human TO pg_monitor;


----------------------------------------------------------------------
-- Activity Overview: monitor.pg_session
----------------------------------------------------------------------
DROP VIEW IF EXISTS monitor.pg_session CASCADE;
CREATE OR REPLACE VIEW monitor.pg_session AS
SELECT coalesce(datname, 'all') AS datname, numbackends, active, idle, ixact, max_duration, max_tx_duration, max_conn_duration
FROM (
         SELECT datname,
                count(*)                                         AS numbackends,
                count(*) FILTER ( WHERE state = 'active' )       AS active,
                count(*) FILTER ( WHERE state = 'idle' )         AS idle,
                count(*) FILTER ( WHERE state = 'idle in transaction'
                    OR state = 'idle in transaction (aborted)' ) AS ixact,
                max(extract(epoch from now() - state_change))
                FILTER ( WHERE state = 'active' )                AS max_duration,
                max(extract(epoch from now() - xact_start))      AS max_tx_duration,
                max(extract(epoch from now() - backend_start))   AS max_conn_duration
         FROM pg_stat_activity
         WHERE backend_type = 'client backend'
           AND pid <> pg_backend_pid()
         GROUP BY ROLLUP (1)
         ORDER BY 1 NULLS FIRST
     ) t;
COMMENT ON VIEW monitor.pg_session IS 'postgres activity group by session';
GRANT SELECT ON monitor.pg_session TO pg_monitor;


----------------------------------------------------------------------
-- Sequential Scan: monitor.pg_seq_scan
----------------------------------------------------------------------
DROP VIEW IF EXISTS monitor.pg_seq_scan CASCADE;
CREATE OR REPLACE VIEW monitor.pg_seq_scan AS
SELECT schemaname                                                        AS nspname,
       relname,
       seq_scan,
       seq_tup_read,
       seq_tup_read / seq_scan                                           AS seq_tup_avg,
       idx_scan,
       n_live_tup + n_dead_tup                                           AS tuples,
       round(n_live_tup * 100.0::NUMERIC / (n_live_tup + n_dead_tup), 2) AS live_ratio
FROM pg_stat_user_tables
WHERE seq_scan > 0
  and (n_live_tup + n_dead_tup) > 0
ORDER BY seq_scan DESC;
COMMENT ON VIEW monitor.pg_seq_scan IS 'table that have seq scan';
GRANT SELECT ON monitor.pg_seq_scan TO pg_monitor;

DROP FUNCTION IF EXISTS monitor.pg_shmem() CASCADE;
CREATE OR REPLACE FUNCTION monitor.pg_shmem() RETURNS SETOF
    pg_shmem_allocations SET search_path = '' AS $$ SELECT * FROM pg_shmem_allocations;$$ LANGUAGE SQL SECURITY DEFINER;
COMMENT ON FUNCTION monitor.pg_shmem() IS 'security wrapper for system view pg_shmem';
REVOKE ALL ON FUNCTION monitor.pg_shmem() FROM PUBLIC;
REVOKE ALL ON FUNCTION monitor.pg_shmem() FROM dbrole_readonly;
REVOKE ALL ON FUNCTION monitor.pg_shmem() FROM dbrole_offline;
GRANT EXECUTE ON FUNCTION monitor.pg_shmem() TO pg_monitor;