为什么postgresql会在磁盘空间足够的情况下提示"磁盘空间不足"？

Question

我在ubuntu 12.0.4上的postgresql中运行此查询:

SELECT t2.p AS prop, t2.o AS obj, COUNT(t2.o) AS num 
FROM "class_Event" AS t1, 
  ((SELECT s,p,o FROM "prop_sliceHasAnomaly")  
   UNION (SELECT s,p,o FROM "prop_eventHasDuration")  
   UNION (SELECT s,p,o FROM "prop_sliceHasEndEvent")  
   UNION (SELECT s,p,o FROM "prop_eventPrecedeInCPU")  
   UNION (SELECT s,p,o FROM "prop_eventFollowInTask")  
   UNION (SELECT s,p,o FROM "prop_topObjectProperty")  
   UNION (SELECT s,p,o FROM "prop_eventDetails")  
   UNION (SELECT s,p,o FROM "prop_switchTo")  
   UNION (SELECT s,p,o FROM "prop_eventIsExecutedOn")  
   UNION (SELECT s,p,o FROM "prop_runningAction")  
   UNION (SELECT s,p,o FROM "prop_anomalyHasSlice")  
   UNION (SELECT s,p,o FROM "prop_eventPrecedeInTrace")  
   UNION (SELECT s,p,o FROM "prop_sliceHasStartEvent")  
   UNION (SELECT s,p,o FROM "prop_eventHasActiveDuration")  
   UNION (SELECT s,p,o FROM "prop_eventFollowInTrace")  
   UNION (SELECT s,p,o FROM "prop_runningTask")  
   UNION (SELECT s,p,o FROM "prop_eventOrdering")  
   UNION (SELECT s,p,o FROM "prop_domain")  
   UNION (SELECT s,p,o FROM "prop_sliceHasFunctionality")  
   UNION (SELECT s,p,o FROM "prop_traceContainsEvent")  
   UNION (SELECT s,p,o FROM "prop_eventHasDurationFromPreviousOccurrence")  
   UNION (SELECT s,p,o FROM "prop_eventPrecedeOccurrence")  
   UNION (SELECT s,p,o FROM "prop_eventPrecedeinTask")  
   UNION (SELECT s,p,o FROM "prop_switchFrom")  
   UNION (SELECT s,p,o FROM "prop_eventEndAt")  
   UNION (SELECT s,p,o FROM "prop_subPropertyOf")  
   UNION (SELECT s,p,o FROM "prop_eventFollowOccurrence")  
   UNION (SELECT s,p,o FROM "prop_eventFollowInCPU")  
   UNION (SELECT s,p,o FROM "prop_subClassOf")  
   UNION (SELECT s,p,o FROM "prop_eventHasDurationToNextOccurrence")  
   UNION (SELECT s,p,o FROM "prop_requestComponent")  
   UNION (SELECT s,p,o FROM "prop_eventStartAt")  
   UNION (SELECT s,p,o FROM "prop_range")  
   UNION (SELECT s,p,o FROM "prop_functionalityHasSlice") ) AS t2 
WHERE t1.s = t2.s 
GROUP BY t2.p, t2.o 
HAVING (COUNT(t2.o) > 1) 

但是我从posgresql出来了这个错误:

ERROR:  could not write block 713 of temporary file: Aucun espace disponible sur le périphérique
HINT:  Perhaps out of disk space?

********** Erreur **********

ERROR: could not write block 713 of temporary file: Aucun espace disponible sur le périphérique
État SQL :53100
Astuce : Perhaps out of disk space?

我已经通过命令检查了足够的磁盘空间 df -h

df -h
df: «/var/lib/lightdm/.gvfs»: Permission non accordée
Sys. de fichiers Taille Utilisé Dispo Uti% Monté sur
/dev/sda5           37G     35G   46M 100% /
udev                16G    4,0K   16G   1% /dev
tmpfs              6,3G    984K  6,3G   1% /run
none               5,0M       0  5,0M   0% /run/lock
none                16G    216K   16G   1% /run/shm
/dev/sda6           37G    411M   35G   2% /opt
/dev/sda9          499G    435G   40G  92% /home
/dev/sda7           37G    4,7G   30G  14% /usr
/dev/sda8           37G    5,1G   30G  15% /usr/local

而且似乎有40GB的可用空间磁盘/home.所以我想知道是什么让这个错误引起的？是否有可以在postgresql配置中设置的磁盘空间限制参数？

有人对此有所了解吗？

Answer 1

它正在写一个临时文件.他们被写入temp_tablespaces,其中:

默认值为空字符串,这会导致在当前数据库的默认表空间中创建所有临时对象.

这意味着它会将临时文件写入你的data_directory.要查看它的位置,请运行SHOW data_directory.我希望你会得到一个结果/var/lib/pgsql/9.3/data/,表明PostgreSQL的数据在/var.

(确保检查一下;如果临时表空间在tempfs上,那么它基本上是一个ramdisk,你应该把它移到别处,因为这对于性能不好,对于大的临时文件也是一个问题.)

在您的系统上,/var是/文件系统的一部分,它已满.因此包含PostgreSQL数据的分区已满,PostgreSQL正确地将此报告为错误.

自由空间的选项包括:

• 确保/tmp删除旧的临时文件

• 从中删除旧的日志文件 /var/log

• 删除旧的PostgreSQL日志.根据OS /发行版,它们的位置有些变化,你没有提到; 他们通常会在/var/log/postgresql/或/var/lib/pgsql/9.3/data/pg_log.重要提示:不要删除PostgreSQL数据目录中的任何其他内容 ; 特别是,pg_xlog它pg_clog是数据库系统的重要组成部分,不得以任何方式删除,移动或更改.

• 删除TRUNCATEPostgreSQL中的表(永久销毁数据)

• DROP在PostgreSQL中ping不需要的索引

• 卸载程序

• ...

但最好的选择是:

• 获得更大的磁盘.