金仓数据库（KingbaseES）日常运维500条命令清单

说明：命令按 V8R6、V9 常见语法整理。不同小版本、兼容模式（Oracle、MySQL）及插件版本下，sys_ 与 pg_ 前缀、集群工具参数可能存在差异；执行前请用 ksql \h、sys_ctl --help、对应工具 --help 核对，破坏性命令务必先在测试环境验证。

一、连接与实例基础管理（1--30）

ksql -U system -d test -h 127.0.0.1 -p 54321：本地或远程连接
ksql -U system -d test -W：连接时提示输入密码
ksql -U system -d test -c "SELECT version();"：执行单条 SQL
ksql -U system -d test -f /home/kes/check.sql：执行脚本文件
ksql -U system -d test -o /home/kes/out.txt -c "SELECT now();"：结果输出到文件
ksql -l：列出实例内所有数据库
ksql -U system -d test -c "\l"：列出数据库及属主、编码
\c dbname：切换数据库
\conninfo：查看当前连接信息
\dn：列出 schema
\dt：列出当前 schema 表
\d+ tablename：查看表结构、存储、分布信息
\di：列出索引
\dv：列出视图
\df：列出函数
\du：列出用户与角色
\dx：列出已安装扩展
\dp：查看对象权限
\timing on：开启 SQL 执行耗时显示
\x：切换扩展显示模式
\encoding：查看客户端字符编码
\q：退出 ksql
sys_ctl status -D $KINGBASE_DATA：查看实例运行状态
sys_ctl start -D $KINGBASE_DATA -l $KINGBASE_DATA/log/start.log：启动实例
sys_ctl stop -D $KINGBASE_DATA -m fast：快速停止实例
sys_ctl restart -D $KINGBASE_DATA -m fast：重启实例
sys_ctl reload -D $KINGBASE_DATA：重载配置文件
sys_ctl promote -D $KINGBASE_DATA：备库提升为主库
sys_isready -h 127.0.0.1 -p 54321：探测实例是否可连接
sys_controldata -D $KINGBASE_DATA：查看控制文件与集群状态

二、日常巡检（31--120）

2.1 实例与版本

SELECT version();
SELECT current_database(), current_user, now();
SELECT sys_postmaster_start_time();
SELECT date_trunc('second', now() - sys_postmaster_start_time()) AS uptime;
SHOW server_version;
SHOW data_directory;
SELECT name, setting FROM sys_settings WHERE name IN ('port','listen_addresses');
SELECT sys_conf_load_time();

2.2 连接与会话

SELECT count(*) FROM sys_stat_activity;
SELECT state, count(*) FROM sys_stat_activity GROUP BY state;
SELECT count(*) FROM sys_stat_activity WHERE state = 'active';
SELECT usename, count(*) FROM sys_stat_activity GROUP BY usename ORDER BY 2 DESC;
SELECT datname, count(*) FROM sys_stat_activity GROUP BY datname;
SELECT pid,usename,datname,client_addr,state,wait_event_type,wait_event,query_start FROM sys_stat_activity WHERE state <> 'idle' ORDER BY query_start;
SELECT setting FROM sys_settings WHERE name = 'max_connections';
SELECT round((SELECT count(*) FROM sys_stat_activity)::numeric * 100 / (SELECT setting::int FROM sys_settings WHERE name='max_connections'), 2) AS conn_pct;
SELECT * FROM sys_stat_activity WHERE backend_type = 'client backend' AND state = 'idle in transaction';
SELECT pid, now()-xact_start AS xact_age, state, query FROM sys_stat_activity WHERE xact_start IS NOT NULL ORDER BY xact_start LIMIT 20;
SELECT pid, now()-state_change AS idle_sec, query FROM sys_stat_activity WHERE state LIKE 'idle%' ORDER BY state_change LIMIT 20;
SELECT * FROM sys_stat_database;

2.3 空间与对象

SELECT datname, sys_size_pretty(sys_database_size(datname)) FROM sys_database ORDER BY sys_database_size(datname) DESC;
SELECT sys_size_pretty(sys_database_size(current_database()));
SELECT spcname, sys_size_pretty(sys_tablespace_size(spcname)) FROM sys_tablespace;
SELECT relname, sys_size_pretty(sys_total_relation_size(oid)) FROM sys_class WHERE relkind='r' ORDER BY sys_total_relation_size(oid) DESC LIMIT 20;
SELECT nspname, sys_size_pretty(sum(sys_total_relation_size(c.oid))) FROM sys_class c JOIN sys_namespace n ON n.oid=c.relnamespace WHERE relkind='r' GROUP BY 1 ORDER BY sum(sys_total_relation_size(c.oid)) DESC;
SELECT relname, n_live_tup, n_dead_tup, last_vacuum, last_autovacuum, last_analyze FROM sys_stat_user_tables ORDER BY n_dead_tup DESC LIMIT 20;
SELECT schemaname, relname, seq_scan, idx_scan, seq_tup_read, n_live_tup FROM sys_stat_user_tables ORDER BY seq_scan DESC LIMIT 20;
SELECT schemaname, relname, indexrelname, idx_scan, idx_tup_read FROM sys_stat_all_indexes ORDER BY idx_scan ASC LIMIT 20;
SELECT * FROM sys_stat_bgwriter;
SELECT * FROM sys_stat_archiver;
SELECT * FROM sys_stat_progress_vacuum;
SELECT * FROM sys_stat_progress_create_index;

2.4 锁与等待

SELECT locktype, mode, granted, count(*) FROM sys_locks GROUP BY 1,2,3 ORDER BY 4 DESC;
SELECT pid, wait_event_type, wait_event, state, query FROM sys_stat_activity WHERE wait_event_type = 'Lock';
SELECT a.pid AS blocked_pid, a.query AS blocked_query, b.pid AS blocking_pid, b.query AS blocking_query FROM sys_stat_activity a JOIN sys_locks l1 ON l1.pid=a.pid AND NOT l1.granted JOIN sys_locks l2 ON l2.locktype=l1.locktype AND l2.relation IS NOT DISTINCT FROM l1.relation AND l2.granted JOIN sys_stat_activity b ON b.pid=l2.pid;
SELECT * FROM sys_locks WHERE NOT granted;
SELECT pid, relation::regclass, mode, granted FROM sys_locks WHERE relation IS NOT NULL;
SELECT count(*) FROM sys_stat_activity WHERE wait_event_type = 'IO';
SELECT wait_event_type, wait_event, count(*) FROM sys_stat_activity WHERE wait_event IS NOT NULL GROUP BY 1,2 ORDER BY 3 DESC;
SELECT * FROM sys_stat_activity WHERE state='active' AND now()-query_start > interval '5 min';
SELECT pid, query FROM sys_stat_activity WHERE backend_xmin IS NOT NULL ORDER BY age(backend_xmin) DESC LIMIT 20;
SELECT datname, age(datfrozenxid) FROM sys_database ORDER BY 2 DESC;
SELECT relname, age(relfrozenxid) FROM sys_class WHERE relkind='r' ORDER BY 2 DESC LIMIT 20;
SELECT count(*) FROM sys_prepared_xacts;
SELECT * FROM sys_prepared_xacts ORDER BY prepared;

2.5 WAL、归档与复制

SELECT slot_name, slot_type, active, restart_lsn, sys_size_pretty(sys_wal_lsn_diff(sys_current_wal_lsn(), restart_lsn)) FROM sys_replication_slots;
SELECT * FROM sys_stat_replication;
SELECT client_addr, state, sync_state, sys_wal_lsn_diff(sent_lsn, replay_lsn) AS replay_lag_bytes FROM sys_stat_replication;
SELECT application_name, state, sys_wal_lsn_diff(sys_current_wal_lsn(), replay_lsn) AS lag_bytes FROM sys_stat_replication;
SELECT sys_last_wal_receive_lsn(), sys_last_wal_replay_lsn();
SELECT sys_current_wal_lsn(), sys_last_wal_replay_lsn();
SELECT sys_wal_lsn_diff(sys_last_wal_receive_lsn(), sys_last_wal_replay_lsn()) AS apply_lag;
SHOW wal_level; SHOW archive_mode; SHOW archive_command;
SELECT name, setting FROM sys_settings WHERE name IN ('wal_level','archive_mode','archive_command','max_wal_size','min_wal_size');
ls -l $KINGBASE_DATA/pg_wal | tail -20：查看在线 WAL 文件
SELECT count(*) FROM sys_ls_waldir();
SELECT archived_count, failed_count, last_archived_wal, last_failed_wal FROM sys_stat_archiver;
du -sh /backup/wal_archive：归档目录容量
SHOW autovacuum; SHOW autovacuum_naptime; SHOW autovacuum_max_workers;

2.6 膨胀与自动清理

SELECT * FROM sys_stat_activity WHERE backend_type LIKE '%autovacuum%';
SELECT relname, n_dead_tup, round(n_dead_tup*100.0/nullif(n_live_tup+n_dead_tup,0),2) AS dead_pct FROM sys_stat_user_tables ORDER BY dead_pct DESC LIMIT 20;
SELECT * FROM sys_stat_progress_vacuum;
SELECT relname, last_autovacuum, autovacuum_count FROM sys_stat_user_tables WHERE last_autovacuum IS NULL AND n_dead_tup > 1000;
SELECT relname FROM sys_stat_user_tables WHERE last_analyze IS NULL;
SELECT count(*) FROM sys_stat_activity WHERE query LIKE '%autovacuum%';
SELECT name, setting, unit FROM sys_settings WHERE name LIKE '%timeout%';

2.7 日志检查

SHOW log_min_duration_statement;
SHOW log_directory; SHOW log_filename; SHOW logging_collector;
tail -n 200 $KINGBASE_DATA/log/*.log
grep -c "ERROR" $KINGBASE_DATA/log/*.log
grep "FATAL" $KINGBASE_DATA/log/*.log | tail -50
grep "deadlock detected" $KINGBASE_DATA/log/*.log | tail -20
grep -i "out of memory" $KINGBASE_DATA/log/*.log | tail -20
df -h：文件系统使用率
df -i：inode 使用率

2.8 主机资源

free -m
top -b -n1 | head -20
iostat -x 1 3
vmstat 1 5
sar -u 1 5
ss -lntp | grep 54321
ps -ef | grep kingbase
uptime
dmesg | tail -50
cat /proc/meminfo
du -sh $KINGBASE_DATA
find $KINGBASE_DATA -type f -size +1G -exec ls -lh {} \;
SELECT * FROM sys_stat_ssl;
SELECT * FROM sys_file_settings;
SELECT * FROM sys_hba_file_rules;

三、性能优化与 SQL 诊断（121--220）

3.1 执行计划

EXPLAIN SELECT * FROM t WHERE id = 1;
EXPLAIN (ANALYZE, BUFFERS) SELECT * FROM t WHERE id = 1;
EXPLAIN (ANALYZE, BUFFERS, VERBOSE, FORMAT JSON) SELECT ...;
EXPLAIN (COSTS OFF) SELECT ...;
EXPLAIN (ANALYZE, TIMING OFF) SELECT ...;

3.2 SQL 统计

SELECT * FROM sys_stat_statements ORDER BY total_exec_time DESC LIMIT 20;
SELECT * FROM sys_stat_statements ORDER BY mean_exec_time DESC LIMIT 20;
SELECT * FROM sys_stat_statements ORDER BY calls DESC LIMIT 20;
SELECT * FROM sys_stat_statements ORDER BY rows DESC LIMIT 20;
SELECT queryid, calls, round(total_exec_time::numeric,2) AS total_ms FROM sys_stat_statements ORDER BY 3 DESC LIMIT 20;
SELECT queryid, query FROM sys_stat_statements WHERE shared_blks_read > 10000;
SELECT sum(blks_hit)*100.0/nullif(sum(blks_hit+blks_read),0) AS cache_hit_pct FROM sys_stat_database;
SELECT datname, blks_hit, blks_read FROM sys_stat_database ORDER BY blks_read DESC;
SELECT * FROM sys_stat_user_tables WHERE seq_scan > idx_scan AND n_live_tup > 10000;
SELECT * FROM sys_stat_all_indexes WHERE idx_scan = 0;
SELECT indexrelname, sys_size_pretty(sys_relation_size(indexrelid)) FROM sys_stat_all_indexes ORDER BY sys_relation_size(indexrelid) DESC LIMIT 20;
SELECT sys_stat_statements_reset();
SELECT sys_stat_reset();
SELECT sys_stat_reset_shared('bgwriter');

3.3 索引与统计信息

CREATE INDEX CONCURRENTLY idx_t_c1 ON t(c1);
REINDEX INDEX CONCURRENTLY idx_t_c1;
REINDEX TABLE t;
ANALYZE t;
ANALYZE VERBOSE t;
VACUUM (VERBOSE, ANALYZE) t;
VACUUM FULL t;
VACUUM (FREEZE, VERBOSE) t;
CLUSTER t USING idx_t_c1;
SELECT * FROM sys_stats WHERE tablename = 't';
SELECT attname, n_distinct, correlation FROM sys_stats WHERE tablename = 't';
ALTER TABLE t ALTER COLUMN c1 SET STATISTICS 500;
SELECT * FROM sys_indexes WHERE tablename = 't';
SELECT reltuples, relpages FROM sys_class WHERE relname = 't';
SELECT relname, relkind, relpages, reltuples FROM sys_class WHERE relname = 't';

3.4 关键参数

SELECT name, setting, unit FROM sys_settings WHERE name LIKE '%cost%';
SHOW random_page_cost; SHOW effective_cache_size;
SHOW work_mem; SHOW maintenance_work_mem;
SHOW max_parallel_workers_per_gather;
SET work_mem = '64MB';
ALTER SYSTEM SET work_mem = '64MB';
SELECT sys_reload_conf();
SHOW shared_preload_libraries;
SHOW effective_io_concurrency;
SHOW shared_buffers; SHOW temp_buffers;
SHOW max_worker_processes; SHOW max_parallel_workers;
SHOW synchronous_commit;
SHOW checkpoint_timeout; SHOW max_wal_size;
SHOW track_io_timing;
SHOW log_min_duration_statement;

3.5 检查点与后台写

CHECKPOINT;
SELECT sys_switch_wal();
SELECT * FROM sys_stat_progress_checkpoint;
SELECT checkpoints_timed, checkpoints_req, buffers_checkpoint, buffers_backend FROM sys_stat_bgwriter;
SELECT * FROM sys_stat_bgwriter;

3.6 锁与阻塞

SELECT * FROM sys_locks WHERE mode = 'ExclusiveLock';
SELECT * FROM sys_locks WHERE locktype = 'advisory';
SELECT * FROM sys_stat_activity WHERE wait_event = 'transactionid';
SELECT pid, query FROM sys_stat_activity WHERE state='active' ORDER BY query_start LIMIT 10;
SELECT sys_terminate_backend(pid);
SELECT sys_cancel_backend(pid);
SELECT * FROM sys_stat_activity WHERE backend_xid IS NOT NULL;
SELECT * FROM sys_locks WHERE relation = 't'::regclass;
SHOW deadlock_timeout; SHOW lock_timeout;

3.7 表与索引访问特征

SELECT * FROM sys_stat_user_functions;
SELECT * FROM sys_stat_user_tables ORDER BY n_tup_upd DESC LIMIT 20;
SELECT * FROM sys_stat_user_tables ORDER BY n_tup_del DESC LIMIT 20;
SELECT relname, n_tup_hot_upd, n_tup_upd FROM sys_stat_user_tables;
SELECT * FROM sys_stat_database_conflicts;
SELECT * FROM sys_stat_activity WHERE backend_type = 'parallel worker';
SELECT relname, sys_size_pretty(sys_relation_size(reltoastrelid)) AS toast_size FROM sys_class WHERE reltoastrelid <> 0 ORDER BY sys_relation_size(reltoastrelid) DESC LIMIT 20;
SELECT relname, sys_size_pretty(sys_table_size(oid)) AS heap, sys_size_pretty(sys_indexes_size(oid)) AS idx FROM sys_class WHERE relkind='r' ORDER BY sys_total_relation_size(oid) DESC LIMIT 20;
SELECT * FROM sys_stat_all_tables WHERE n_live_tup = 0 AND n_dead_tup > 0;

3.8 会话与超时控制

SET statement_timeout = '30s';
SET lock_timeout = '5s';
SET idle_in_transaction_session_timeout = '10min';
SET idle_session_timeout = '30min';
SHOW statement_timeout; SHOW idle_in_transaction_session_timeout;
ALTER SYSTEM SET idle_in_transaction_session_timeout = '10min';

3.9 临时文件与 IO

SET log_temp_files = 0;
SHOW temp_file_limit;
SELECT datname, temp_files, temp_bytes FROM sys_stat_database ORDER BY temp_bytes DESC;
SELECT * FROM sys_stat_activity WHERE wait_event_type = 'IO';
SELECT * FROM sys_stat_activity WHERE wait_event_type = 'LWLock';

3.10 优化辅助

SELECT * FROM sys_prepared_statements;
SELECT * FROM sys_cursors;
SELECT name, statement, prepare_time FROM sys_prepared_statements ORDER BY prepare_time;
EXPLAIN (ANALYZE, BUFFERS) SELECT ... /* 验证索引改写效果 */;
SELECT * FROM sys_stat_statements WHERE query ILIKE '%seq scan%';
SELECT queryid, query FROM sys_stat_statements WHERE total_exec_time > 60000;
SELECT * FROM sys_stat_statements ORDER BY shared_blks_hit DESC LIMIT 10;
SELECT * FROM sys_stat_statements ORDER BY wal_bytes DESC LIMIT 10;
SELECT * FROM sys_stat_statements ORDER BY temp_blks_written DESC LIMIT 10;
SELECT sys_size_pretty(sum(temp_bytes)) FROM sys_stat_database;
CREATE EXTENSION IF NOT EXISTS pg_hint_plan;（按版本核对扩展名）
SELECT /*+ IndexScan(t idx_t_c1) */ * FROM t WHERE c1 = 1;（需 hint 插件）
SET enable_seqscan = off;
SET enable_nestloop = off;
SET enable_hashjoin = on;
SET max_parallel_workers_per_gather = 4;
SELECT * FROM sys_settings WHERE name IN ('jit','jit_above_cost','parallel_setup_cost');

四、备份与恢复（221--325）

4.1 逻辑备份

sys_dump -U system -d test -f /backup/test.sql
sys_dump -U system -d test -Fc -f /backup/test.dump
sys_dump -U system -d test -Fp -f /backup/test.sql
sys_dump -U system -d test -Ft -f /backup/test.tar
sys_dump -U system -d test -Fd -j 4 -f /backup/test_dir
sys_dump -U system -d test -n public -f /backup/public.sql
sys_dump -U system -d test -t t -f /backup/t.sql
sys_dump -U system -d test -T t -f /backup/except_t.sql
sys_dump -U system -d test --inserts -f /backup/test_inserts.sql
sys_dump -U system -d test --column-inserts -f /backup/test_col.sql
sys_dump -U system -d test --no-owner -f /backup/test.sql
sys_dump -U system -d test --no-privileges -f /backup/test.sql
sys_dump -U system -d test --clean --if-exists -f /backup/test.sql
sys_dump -U system -d test --schema-only -f /backup/schema.sql
sys_dump -U system -d test --data-only -f /backup/data.sql
sys_dump -U system -d test -Z 9 -Fc -f /backup/test.dump
sys_dump -U system -d test --verbose -f /backup/test.sql
sys_dump -U system -d test -j 4 -Fd -f /backup/test_par
sys_dumpall -U system -g -f /backup/globals.sql
sys_dumpall -U system -f /backup/all.sql
sys_dumpall -U system -r -f /backup/roles.sql
sys_dumpall -U system -s -f /backup/all_schema.sql
sys_dump -U system -d test -b -f /backup/with_blob.sql
sys_dump -U system -d test --exclude-table-data=t_log -f /backup/test.sql
sys_dump -U system -d test --exclude-schema=history -f /backup/test.sql
sys_dump -U system -d test --section=pre-data -f /backup/pre.sql
sys_dump -U system -d test --serializable-deferrable -f /backup/test.sql

4.2 逻辑恢复

sys_restore -U system -d test /backup/test.dump
sys_restore -U system -C -d postgres /backup/test.dump
sys_restore -l /backup/test.dump
sys_restore -L /backup/toc.list -d test /backup/test.dump
sys_restore -j 4 -d test /backup/test_dir
sys_restore -t t -d test /backup/test.dump
sys_restore -n public -d test /backup/test.dump
sys_restore --clean --if-exists -d test /backup/test.dump
sys_restore --data-only -d test /backup/test.dump
sys_restore --schema-only -d test /backup/test.dump
ksql -U system -d test -f /backup/test.sql
createdb -U system -T template0 test_new
CREATE DATABASE test_new TEMPLATE test;
dropdb -U system test_new
DROP DATABASE IF EXISTS test_new;

4.3 表级与数据交换

COPY t TO '/backup/t.csv' WITH (FORMAT csv, HEADER true);
COPY t FROM '/backup/t.csv' WITH (FORMAT csv, HEADER true);
\copy t TO '/backup/t.csv' CSV HEADER（ksql 元命令）
\copy t FROM '/backup/t.csv' CSV HEADER
COPY (SELECT * FROM t WHERE dt='2026-11-01') TO '/backup/t_1101.csv' CSV HEADER;
COPY t TO PROGRAM 'gzip > /backup/t.csv.gz' CSV HEADER;
SELECT * FROM sys_stat_progress_copy;（按版本核对视图）

4.4 物理备份

sys_basebackup -D /backup/base -Fp -Xs -P -R
sys_basebackup -D /backup/base -Ft -z -P
sys_basebackup -D /backup/base -Fp -Xs -P --checkpoint=fast
sys_basebackup -D /backup/base -X stream -c fast -P
sys_basebackup -D /backup/base -l "full_20261110" -P
sys_basebackup -h standby_host -D /backup/base -Xs -P
sys_basebackup -D /backup/base -Fp -Xs -P -R -S slot_name

4.5 备份工具 sys_rman

sys_rman init
sys_rman --config=/backup/backup.info --stanza=kes backup
sys_rman --config=/backup/backup.info --stanza=kes backup --type=full
sys_rman --config=/backup/backup.info --stanza=kes backup --type=incr
sys_rman --config=/backup/backup.info --stanza=kes backup --type=diff
sys_rman --config=/backup/backup.info --stanza=kes check
sys_rman --config=/backup/backup.info --stanza=kes info
sys_rman --config=/backup/backup.info --stanza=kes verify
sys_rman --config=/backup/backup.info --stanza=kes restore
sys_rman --config=/backup/backup.info --stanza=kes expire
sys_rman --config=/backup/backup.info --stanza=kes archive-push
sys_rman --config=/backup/backup.info --stanza=kes archive-get
sys_rman --config=/backup/backup.info --stanza=kes stanza-create
sys_rman --config=/backup/backup.info --stanza=kes stanza-upgrade
sys_rman --config=/backup/backup.info --stanza=kes delete --set=...
sys_rman --config=/backup/backup.info --stanza=kes backup --backup-standby
sys_rman --config=/backup/backup.info --stanza=kes info --output=json

4.6 备份脚本

sys_backup.sh init
sys_backup.sh start
sys_backup.sh stop
sys_backup.sh status
crontab -e：配置定时备份任务
0 2 * * * /home/kes/sys_backup.sh start >/dev/null 2>&1：每日 2 点备份
sys_dump -Fc -f /backup/test_$(date +%F).dump test：按日期命名备份
sys_restore -l /backup/test_20261110.dump | head -20：备份可读性校验
md5sum /backup/test_20261110.dump：备份文件完整性校验

4.7 时间点恢复（PITR）

restore_command = 'cp /backup/wal_archive/%f %p'
recovery_target_time = '2026-11-10 09:30:00'
recovery_target_lsn = '0/1A2B3C4D'
recovery_target_xid = '1234567'
recovery_target_name = 'before_upgrade'
recovery_target_action = 'promote'
touch $KINGBASE_DATA/recovery.signal
SELECT sys_create_restore_point('before_upgrade');
SELECT sys_switch_wal();
ls -l /backup/wal_archive | tail -20
sys_waldump /backup/wal_archive/000000010000000000000001 | head -50
恢复完成后执行 ANALYZE; 刷新统计信息
恢复后执行 sys_vacuumdb --all --analyze（工具名按版本核对）
恢复到新实例：sys_basebackup + 独立端口启动
恢复后核对 SELECT count(*) FROM t;
恢复后核对序列当前值 SELECT last_value FROM seq_name;
恢复后核对复制槽状态
恢复后重建索引统计
恢复演练记录表结构与数据量比对
备份保留策略：find /backup -name "*.dump" -mtime +7 -delete（谨慎）
异地备份：rsync -avz /backup/ backup_host:/backup/
备份文件加密：gpg -c /backup/test.dump
备份目录权限检查：ls -ld /backup

五、高可用与主备复制（326--385）

5.1 repmgr 集群管理

repmgr primary register
repmgr standby register
repmgr cluster show
repmgr cluster crosscheck
repmgr cluster event --limit 20
repmgr service status
repmgr standby promote
repmgr standby switchover
repmgr node rejoin
repmgr cluster cleanup
repmgr standby clone -h primary_host -D $KINGBASE_DATA
repmgr node status
repmgr witness register
repmgr cluster matrix
repmgr daemon start

5.2 复制参数与状态

SHOW synchronous_commit;
SHOW synchronous_standby_names;
ALTER SYSTEM SET synchronous_standby_names = 'standby1';
SELECT * FROM sys_stat_replication;
SELECT client_addr, state, sync_state, sent_lsn, replay_lsn FROM sys_stat_replication;
SELECT slot_name, slot_type, active FROM sys_replication_slots;
SELECT sys_create_physical_replication_slot('slot_name');
SELECT sys_drop_replication_slot('slot_name');
SELECT sys_replication_slot_advance('slot_name', sys_current_wal_lsn());
SHOW wal_level; SHOW max_wal_senders; SHOW hot_standby;
SELECT sys_is_in_recovery();
SELECT sys_last_wal_receive_lsn(), sys_last_wal_replay_lsn();
SELECT sys_wal_lsn_diff(sys_last_wal_receive_lsn(), sys_last_wal_replay_lsn());
SHOW primary_conninfo;
SHOW restore_command;
SHOW wal_keep_segments;
SELECT * FROM sys_stat_database_conflicts;

5.3 切换与恢复

sys_ctl promote -D $KINGBASE_DATA
sys_rewind --target-pgdata=$KINGBASE_DATA --source-server="host=primary port=54321 user=system dbname=test"
sys_rewind --target-pgdata=$KINGBASE_DATA --source-pgdata=/backup/base
SELECT sys_create_restore_point('switchover_point');
SELECT sys_switch_wal();
切换前检查：SELECT * FROM sys_stat_replication;
切换前检查：SELECT slot_name, active FROM sys_replication_slots;
切换后校验：SELECT sys_is_in_recovery();
切换后校验：SELECT count(*) FROM 关键业务表;
回切前检查主备延迟为 0
回切命令：repmgr standby switchover
双主检查：主备同时 SELECT sys_is_in_recovery();
脑裂检查：比对 sys_current_wal_lsn()
VIP 切换：ip addr add 10.0.0.10/24 dev eth0
VIP 摘除：ip addr del 10.0.0.10/24 dev eth0
systemctl status keepalived
ip -4 addr show dev eth0
ping -c 4 standby_host
chronyc sources -v
ntpq -p
date：主备时间一致性检查

5.4 逻辑复制

CREATE PUBLICATION pub_all FOR ALL TABLES;
CREATE PUBLICATION pub_t FOR TABLE t;
CREATE SUBSCRIPTION sub_t CONNECTION 'host=... dbname=...' PUBLICATION pub_t;
SELECT * FROM sys_publication;
SELECT * FROM sys_subscription;
SELECT * FROM sys_stat_subscription;
ALTER SUBSCRIPTION sub_t REFRESH PUBLICATION;

六、故障排查（386--450）

6.1 连接类

SELECT count(*), state FROM sys_stat_activity GROUP BY state;
SELECT setting FROM sys_settings WHERE name='max_connections';
ALTER SYSTEM SET max_connections = 1000;（需重启）
SHOW listen_addresses;
SELECT * FROM sys_hba_file_rules;
grep "password authentication failed" $KINGBASE_DATA/log/*.log | tail -20
grep "no pg_hba.conf entry" $KINGBASE_DATA/log/*.log | tail -20
ss -lntp | grep 54321
sys_isready -h 127.0.0.1 -p 54321
ps -ef | grep postgres

6.2 锁与事务类

SELECT * FROM sys_locks WHERE NOT granted;
阻塞树查询：见第 65 条
grep "deadlock detected" $KINGBASE_DATA/log/*.log | tail -10
SHOW deadlock_timeout;
SELECT sys_terminate_backend(pid) FROM sys_stat_activity WHERE state='idle in transaction' AND now()-state_change > interval '30 min';
SELECT pid, now()-xact_start FROM sys_stat_activity WHERE xact_start IS NOT NULL ORDER BY 2 DESC LIMIT 10;
SELECT * FROM sys_prepared_xacts;
SELECT sys_advisory_unlock_all();

6.3 膨胀与 XID

SELECT relname, n_dead_tup FROM sys_stat_user_tables ORDER BY n_dead_tup DESC LIMIT 20;
VACUUM (VERBOSE, ANALYZE) t;
VACUUM FULL t;
SELECT datname, age(datfrozenxid) FROM sys_database ORDER BY 2 DESC;
SELECT relname, age(relfrozenxid) FROM sys_class ORDER BY 2 DESC LIMIT 20;
VACUUM FREEZE t;
单用户模式维护：sys_single -D $KINGBASE_DATA（工具名按版本核对）

6.4 WAL 与归档

SELECT * FROM sys_stat_archiver;
SHOW archive_command;
sh -c "$(ksql -At -c 'SHOW archive_command' | sed 's/%p/\/tmp\/test_wal/g; s/%f/test_wal/g')"
df -h $KINGBASE_DATA/pg_wal
SELECT count(*) FROM sys_ls_waldir();
SELECT slot_name, sys_size_pretty(sys_wal_lsn_diff(sys_current_wal_lsn(), restart_lsn)) FROM sys_replication_slots;
SELECT sys_drop_replication_slot('inactive_slot');

6.5 复制异常

SELECT * FROM sys_stat_replication;
SELECT sys_last_wal_receive_lsn(), sys_last_wal_replay_lsn();
SELECT * FROM sys_stat_database_conflicts;
SHOW hot_standby_feedback;
SHOW max_standby_streaming_delay;
SELECT sys_is_in_recovery();
tail -n 100 $KINGBASE_DATA/log/*.log | grep -i "replication"

6.6 资源类

dmesg | grep -i "out of memory"
free -m
top -b -n1 -o %CPU | head -20
top -b -n1 -o %MEM | head -20
iostat -x 1 3
pidstat -d 1 5
SELECT * FROM sys_stat_activity WHERE wait_event_type='IO';
SELECT * FROM sys_stat_statements ORDER BY total_exec_time DESC LIMIT 10;
SELECT wait_event_type, wait_event, count(*) FROM sys_stat_activity GROUP BY 1,2 ORDER BY 3 DESC;
SELECT datname, temp_files, temp_bytes FROM sys_stat_database ORDER BY 3 DESC;
SHOW log_temp_files;

6.7 空间与文件

df -h
df -i
du -sh $KINGBASE_DATA/*
SELECT spcname, sys_size_pretty(sys_tablespace_size(spcname)) FROM sys_tablespace;
lsof | grep deleted | head -20
ulimit -n
cat /proc/sys/fs/file-max
cat /proc/sys/vm/overcommit_memory
cat /proc/sys/vm/swappiness

6.8 启动与崩溃

sys_ctl status -D $KINGBASE_DATA
sys_controldata -D $KINGBASE_DATA | grep -i state
tail -n 200 $KINGBASE_DATA/log/*.log
ls -l $KINGBASE_DATA/postmaster.pid
sys_resetwal -D $KINGBASE_DATA（高危，仅崩溃后按手册使用）
REINDEX TABLE t;：索引损坏重建

七、空间、对象与日常维护（451--478）

VACUUM;
VACUUM ANALYZE;
VACUUM FULL;
ALTER SYSTEM SET autovacuum_vacuum_scale_factor = 0.1;
ALTER TABLE t SET (autovacuum_vacuum_scale_factor = 0.05);
ANALYZE;
REINDEX DATABASE test;
REINDEX INDEX CONCURRENTLY idx_t_c1;
CLUSTER t USING idx_t_c1;
TRUNCATE TABLE t;
ALTER TABLE t SET (fillfactor = 80);
CREATE TABLE t_p (id int, dt date) PARTITION BY RANGE (dt);
CREATE TABLE t_p_202611 PARTITION OF t_p FOR VALUES FROM ('2026-11-01') TO ('2026-12-01');
ALTER TABLE t_p ATTACH PARTITION t_p_202610 FOR VALUES FROM ('2026-10-01') TO ('2026-11-01');
ALTER TABLE t_p DETACH PARTITION t_p_202601;
CREATE TABLESPACE ts_data LOCATION '/data/kes/ts_data';
ALTER TABLE t SET TABLESPACE ts_data;
ALTER INDEX idx_t_c1 SET TABLESPACE ts_data;
CREATE SEQUENCE seq_t START 1 INCREMENT 1 CACHE 20;
SELECT last_value, is_called FROM seq_t;
ALTER SEQUENCE seq_t RESTART WITH 10000;
REFRESH MATERIALIZED VIEW CONCURRENTLY mv_t;
SELECT * FROM sys_matviews;
SELECT * FROM sys_sequences;
SELECT * FROM sys_views WHERE schemaname='public';
SELECT * FROM sys_tables WHERE schemaname='public';
SELECT relname, n_live_tup FROM sys_stat_user_tables ORDER BY n_live_tup DESC LIMIT 20;
SELECT * FROM sys_stat_progress_vacuum;

八、安全、权限与审计（479--500）

CREATE USER app_user WITH PASSWORD '******';
ALTER USER app_user WITH PASSWORD '******';
CREATE ROLE read_role NOLOGIN;
GRANT CONNECT ON DATABASE test TO app_user;
GRANT USAGE ON SCHEMA public TO app_user;
GRANT SELECT ON ALL TABLES IN SCHEMA public TO read_role;
REVOKE DELETE ON t FROM app_user;
ALTER DEFAULT PRIVILEGES IN SCHEMA public GRANT SELECT ON TABLES TO read_role;
SELECT * FROM sys_roles;
SELECT * FROM sys_user;
SELECT grantee, table_schema, table_name, privilege_type FROM information_schema.table_privileges WHERE grantee='app_user';
ALTER ROLE app_user VALID UNTIL '2027-01-01';
SHOW password_encryption;
SHOW ssl;
ALTER SYSTEM SET ssl = on;
SELECT * FROM sys_stat_ssl;
SHOW audit_enabled;（审计参数按版本核对）
SELECT * FROM sys_audit_log ORDER BY 1 DESC LIMIT 20;（审计视图按版本核对）
ALTER SYSTEM SET log_connections = on;
ALTER SYSTEM SET log_disconnections = on;
ALTER SYSTEM SET log_statement = 'ddl';
SELECT * FROM sys_settings WHERE name LIKE 'audit%';

使用提示

巡检建议每日执行第二类中"实例、会话、锁、长事务、WAL 归档、autovacuum、日志"七组，形成基线。
性能问题优先看 sys_stat_statements + EXPLAIN (ANALYZE, BUFFERS) + 等待事件三件套，再决定索引、参数或 SQL 改写。
备份恢复务必区分逻辑备份（sys_dump/sys_restore）与物理备份（sys_basebackup/sys_rman），PITR 依赖归档 WAL 连续可用。
主备环境以 repmgr 和 sys_stat_replication 为核心观测点，切换前确认延迟为 0、复制槽正常。
标有"按版本核对"的命令涉及插件、视图或工具参数差异，执行前请以目标版本帮助信息为准。