GaussDB(DWS) 日常维护命令

在日常使用GaussDB(DWS) 过程中，会遇到各种各样的问题，通过熟练的掌握常用的维护命令和问题定位方法，可以使我们提高问题定位效率，快速解决问题。根据以往的经验，将常用的操作命令分成了以下三个部分。在实际使用的过程中可能还需要掌握其它更多的命令，本文仅列举了部分命令。

1、日常维护DB命令

• 会话查杀

• select pg_terminate_backend(procpid); --杀掉会话

  select pg_cancel_backend(procpid); --取消正在执行的语句

• 主备切换命令

  将DN备实例切换为主实例。假设备实例所在主机plat1，路径为"/gaussdb/data/data_dnS1"。

  gs_om -t switch -h plat1 -D /gaussdb/data/data_dnS1

  参数q表示快速切换，nodeid为需要升主的备实例所在节点ID，/srv/BigData/mppdb/data2为备DN或GTM的数据目录。

  cm_ctl switchover -n nodeid -D /srv/BigData/mppdb/data2 -q

• 查看表分布情况

  select getdistributekey('item');

  SELECT n.nspname ,c.relname ,getdistributekey(c.oid) FROM pg_catalog.pg_class c LEFT JOIN pg_catalog.pg_namespace n ON n.oid = c.relnamespace WHERE n.nspname <> 'pg_catalog' AND n.nspname <> 'information_schema' AND n.nspname <> 'cstore' AND c.relkind = 'r' ORDER BY 1,2; --查找多个表的分布列信息

  select pg_size_pretty(pg_table_size('public.item'));

  select table_skewness('inventory');

• 查询审计日志

  select * from pgxc_query_audit('2020-07-16 10:36:05','2020-07-16 12:36:05') where username!='omm';

  SELECT * FROM pg_catalog.pgxc_query_audit_ext ('2021-09-01 19:15:00',current_timestamp) where username = 'xxx' and audit_type = 'user_login'; --查看审计日志

• 查询pooler池

  select node_name, in_use, count() from pg_pooler_status group by node_name, in_use order by 3;
  select database,user_name,in_use,count( ) from pg_pooler_status group by 1, 2, 3 order by 4;

  clean connection to all for database xxx;

• 根据filenode 查找对应的物理表

  select oid, * from pg_class where reltoastrelid = (select oid from pg_class where relfilenode = 103892072);

• 查看内存使用情况

  select * from pgxc_total_memory_detail where memorytype = 'dynamic_used_memory' order by 3 desc;

  select split_part(pv_session_memory_detail.sessid,'.',2),sum(totalsize),count(*) from pv_session_memory_detail group by split_part(pv_session_memory_detail.sessid,'.',2) order by sum(totalsize) desc;

  select sessid, contextname, level,parent, pg_size_pretty(totalsize) as total ,pg_size_pretty(freesize) as freesize, pg_size_pretty(usedsize) as usedsize, datname,query_id, query from pv_session_memory_detail a , pg_stat_activity b where split_part(a.sessid,'.',2) = b.pid and query_id = '76561193666355359' order by totalsize desc limit 100;

• 查询等待视图

  select query_start, state_change, waiting, enqueue, state, a.query_id, substr(replace(query, chr(10), ' '), 0, 10), node_name,thread_name,tid,lwtid,ptid,tlevel,smpid,wait_status,wait_event from pgxc_stat_activity a, pgxc_thread_wait_status b where state = 'active' and a.query_id = b.query_id and a.query_id <> 0;

  select node_name, wait_status, count(*) from pgxc_thread_wait_status group by node_name, wait_status order by 3 desc;

  select nodename,username,application_name, start_time, max_peak_memory , queryid, substr(query,1, 10), substring(warning from 'Statistic Not Collect') as warning from wlm_session where warning like '%Statistic Not Collect%' and application_name = 'Data Studio' order by max_peak_memory desc;

• 查看活跃会话信息

  select coorname, usename, datname, enqueue , count(*) from pgxc_stat_activity where usename <> 'omm' and state = 'active' group by coorname, usename, datname, enqueue ;

  select coorname, usename, client_addr, sysdate - query_start as dur, enqueue, query_id, replace(query, chr(10), ' ') from pgxc_stat_activity where usename!= 'omm' and state = 'active' order by coorname, dur desc;

  SELECT coorname, usename ,client_addr ,sysdate - query_start AS dur ,query_id ,substr(replace(query, chr(10), ' '), 0, 100) FROM pgxc_stat_activity WHERE usename != 'omm' AND STATE = 'active' ORDER BY dur DESC;

• 集群负载管理相关视图查询

  select usename,enqueue,datname,status,attribute,count(),sum(statement_mem) from pg_session_wlmstat group by 3,1,2,4,5 order by 1,3,4,5 ;
  select usename,processid,threadid,priority,attribute,lane,enqueue,status,block_time,elapsed_time,statement_mem from pg_session_wlmstat where usename='usr1';
  select * from pg_stat_get_workload_struct_info();
  select count() from pg_stat_get_wlm_realtime_session_info(NULL);

• 查找删除复制槽

  select * from pg_get_replication_slots(); -- 查找复制槽

  select pg_drop_replication_slot('dn_6004'); --删除复制槽信息

• 查看集群事务信息

  select * from pg_prepared_xacts; --查看残留事务

  select * from pgxc_prepared_xacts; --查看全局残留事务视图

  select * from pg_running_xacts; --查看运行时事务情况

  select * from pgxc_running_xacts; --查看集群运行事务情况

• 集群启停

  checkpoint;

  cm_ctl stop -mi

  cm_ctl start -mi

• 数据文件和日志解析

  pg_xlogdump 000000010000000000000002 -z

  pg_xlogdump 000000010000000000000004 -n

  pagehack -f pg_filenode.map -t filenode_map

  pagehack -f 16502 -t heap

2 常用GUC参数设置

• 检查active sql配置

  show use_workload_manager;

  show enable_control_group;

  show enable_resource_record;

  show enable_resource_track;

  show resource_track_level;

  show resource_track_duration;

  show resource_track_cost;

• 打开 TOPSQL功能

  set use_workload_manager = on;

  set enable_control_group = on;

  set enable_resource_record = on;

  set resource_track_level = query;

• 修改收集统计信息的比例

  set default_statistics_target = -10;

  analzye public.customer;

• 打开集群DEBUG2 级别的日志

  set log_min_messages=debug2;

  set logging_module='on(ALL)';

• 查看页面上所有元组的事务信息

  start transaction read only;

  set enable_show_any_tuples = true;

  set enable_indexscan = off;

  set enable_bitmapscan = off;

  select xmin,xmax,pgxc_is_committed(xmin),pgxc_is_committed(xmax),oid,* from pg_class where relname='表名' ;

• 优化器相关参数，通过调整参数干预估算模型

  cost_model_version;

  cost_param

  join_num_distinct

  qual_num_distinct

• 取消集群只读设置只读级别

  gs_guc reload -Z coordinator -Z datanode -N all -I all -c "default_transaction_read_only=off"

  gs_guc reload -Z coordinator -Z datanode -N all -I all -c "datastorage_threshold_value_check=95"

3 OS相关命令

• 查看进程相关信息
  ps -eo pid,lstart,etime,cmd | grep gaussdb;date
  strace -p 47148 -r -T -o strace.log
• 查看透明大页信息
  cat /sys/kernel/mm/transparent_hugepage/enabled

  never\]表示未打开

  ps -ef | grep gsql | grep -v grep | awk '{print $2}' | xargs kill -9
• 查看CPU占用高的线程
  top -b -p 39450 -H -n 1|head -100
• sar命令
  sar -r 5 4 输出物理内存和虚拟内存的统计信息
  sar -B 5 5 分页统计
  sar -u 3 5 显示CPU使用信息
  sar -b 3 5 磁盘IO信息
  sar -n DEV 2 3 网络流量信息
• 网络问题定位常用命令
  netstat --anop|grep "on ("| sort --rnk 3|head -50
  netstat -naop | grep 54321 --查看端口被占用。
  netstat -anop | awk '{print $4}' | grep ip|sort|uniq -c|grep " 1 "|wc -l --随机端口不足
  ping -s 8192 -I eth0 dest_ip --对端IP是否可达
• 查看raid 卡缓存策略 Write through，IO性能比WriteBack 要慢。
  /opt/MegaRAID/MegaCli/MegaCli64 -LDinfo -Lall --aAll

转载：华为云论坛