Pivotal Knowledge Base

Follow

Large table insert / select fills up OS disk space

Problem

  • Large table insert / selects fills up segment disk space.
  • Grenplum database runs out of space while inserting into table using select or any query that create workfiles etc.
  • Greenplum server reports 100% usage on mount points say /data1 and /data2

Cause

All servers have OS parameter set to allocsize=1g in "/etc/fstab."

Solution

The fix is to edit all the "/etc/fstab" in Greenplum cluster and make the allocsize to 16m.

To do allocsize changes in vfstab follow these steps:

-- Stop the Greenplum database:

gpstop -a

-- Change the following entries in /etc/fstab on all DCA nodes.

LABEL=/data1  /data1 xfs  noatime,inode64,allocsize=1g 0 0
LABEL=/data2  /data2 xfs  noatime,inode64,allocsize=1g 0 0

TO

LABEL=/data1 /data1  xfs  noatime,inode64,allocsize=16m 0 0
LABEL=/data2 /data2  xfs  noatime,inode64,allocsize=16m 0 0

-- Umount and mount file system:

a. # umount /data1
   # umount /data2
b. # mount /data1
   # mount /data2
c. For mdw and smdw:
   # umount /data
   # mount /data

-- Start the database.

Comments

Powered by Zendesk