Oracle Exadata Hardware
 
Exadata Database Machine X2-8
 

PLA-Support>Eng Systems>Exadata/ODA/SSC>Oracle Exadata>DB: Exadata_EST
 

Applies to:

Symptoms

The Cell Server and Database Compute Node Linux imagers create ext3 filesystems without the 'dir_index' and 'resize_inode' options when a node is freshly imaged.

The dir_index option causes directory entries to be indexed which speeds up access to directories with large numbers of files.  On the Database Machine, this is most likely to affect access to the ASM and RDBMS audit and trace directories.

The resize_inode option reserves extra inode space to accommodate larger resize operations on ext3 filesystems.  It is still possible to resize an ext3 filesystem that was not created with the resize_inode option, but the amount by which the filesystem can be grown is lower.

For Example:

dir_index is usually configured as a default on Linux, but none of the filesystems on the Exadata DBM have it configured:

Cause

<Bug 14504499> "LINUX IMAGES MISSING DEFAULT EXT3 FS OPTIONS DIR_INDEX AND RESIZE_INODE"

REDISCOVERY INFORMATION:

tune2fs -l on a file system partition (e.g. /dev/VGExaDb/LVDbSys1 ) does not show dir_index and resize_inode in the Filesystem features line.

 

Solution

<Bug 14504499> is fixed in Exadata systems that have been imaged with the 11.2.3.2.1 version. 

If you are running an earlier version, you can use the following workaround:

Steps to enable dir_index option on Compute Node:

1. Add the dir_index option to both filesystems:

$ tune2fs -O +dir_index /dev/mapper/VGExaDb-LVDbSys1
$ tune2fs -O +dir_index /dev/mapper/VGExaDb-LVDbOra1

2. Verify that option has been added:

$ tune2fs -l /dev/mapper/VGExaDb-LVDbSys1
$ tune2fs -l /dev/mapper/VGExaDb-LVDbOra1

3. Reboot the system using diagnostics.iso. The ISO is located at /opt/oracle.SupportTools/diagnostics.iso. Use scp and copy it and use iLO interface to boot the server with this ISO image.

4. Make sure the filesystems Sys1 and Ora1 are not mounted anywhere. There are chances it get mounted automatically on /mnt/cell. Please take this precaution as running e2fsck on mounted filesystems might corrupt it. You can check it by running command 'mount'.  If it is mounted use 'umount <mount point path>' to unmount it.  (Normally, if the filesystem is mounted, e2fsck command gives a warning and prompts to press y/n before proceeding.)

5. Initiate indexing of the file system directories for each volume where dir_index has been added:

$ e2fsck -D /dev/mapper/VGExaDb-LVDbSys1
$ e2fsck -D /dev/mapper/VGExaDb-LVDbOra1

6. Boot the system normally once indexing has been completed.

The resize_inode option cannot be added to a filesystem which was created with this option missing. Such a filesystem cannot be resized beyond the maximum size reported by resize2fs.

If a larger size is needed, the only workaround is to backup the contents of the filesystem, recreate it with the desired size and restore the backup.

Steps to recreate the filesystem: 

1. Back up the contents.

2. Create the file /etc/mke2fs.conf containing the following:

[defaults]
base_features = sparse_super,filetype,resize_inode,dir_index
blocksize = 4096
inode_ratio = 8192

[fs_types]
small = {
blocksize = 1024
inode_ratio = 4096
}
floppy = {
blocksize = 1024
}
news = {
inode_ratio = 4096
}
largefile = {
inode_ratio = 1048576
}
largefile4 = {
inode_ratio = 4194304
}

3. Run the steps in section "7.25 Resizing LVM Partitions" of chapter "7 Maintaining Oracle Exadata Database Machine and Oracle Exadata Storage Expansion Rack" of Oracle® Exadata Database Machine Owner's Guide
11g Release 2 (11.2)

4. Substitute the "resize2fs" step from the Owner's Guide with:

# mkfs.ext3

5. Restore the backup

References

This solution has no attachment