I suspect your table has index, or checkpoint_segments is small and lead PG do checkpoint frequently. If the table has index or constraint, drop it and copy it ,after copy finished, do create index or constraint again. If checkpoint_segments is small, enlarge it. And also you can turn fsync off when you do copy, after finish, turn it on again. And also you can enlarge maintenance_work_mem. If you take above, time cost will down significantly. 莫建祥 阿里巴巴软件(上海)有限公司 研发中心－IM服务端开发部 联系方式:86-0571-85022088-13072 贸易通ID:jaymo 淘宝ID:jackem 公司网站:www.alisoft.com wiki:http://10.0.32.21:1688/confluence/pages/viewpage.action?pageId=10338 -----邮件原件----- 发件人: pgsql-performance-owner@xxxxxxxxxxxxxx [mailto:pgsql-performance-owner@xxxxxxxxxxxxxx] 代表 Ryan Hansen 发送时间: 2008年9月11日 1:14 收件人: pgsql-performance@xxxxxxxxxxxxxx 主题: Re: [PERFORM] Improve COPY performance for large data sets NEVERMIND!! I found it. Turns out there was still a constraint on the table. Once I dropped that, the time went down to 44 minutes. Maybe I am an idiot after all. :) -Ryan
