Exadata storage reconfiguration with empty DGs

Changing the size of your disk groups (DG) is quite easy if the disk groups are empty. if the total amount of space used in all DGs is less then the smallest DG it’s possible to change the size. it’s like playing towers of Hanoi, you must move your data several times. This is not described here. Here you find instructions to change the size with empty DGs.cell_physicaldisks

The layout of a cell is always the same only the size and amount of disks or flash devices is different.
red = DATA
yellow = RECO
blue=DBFS_DG
green=root filesystem of the cell

The root filesystem of the cell is located on disk 00 and 01. As you can see it’s not useful to change the DBFS_DG (sometimes called CRS or GRID), because of the blend. Therefore it’s not necessary to relocate the OCR of voting files.

The first step is to umount and drop the disk groups as owner of the GI in my case the user grid. If you want to change also the names of the disk groups you should remove it from the cluster as first step. Otherwise you can reuse the resources.
Please check the caching policy and DG attributes!

grid@db01# srvctl stop diskgroup -diskgroup data
grid@db01# srvctl stop diskgroup -diskgroup reco
sql@ASM> alter diskgroup data mount restircted;
sql@ASM> drop diskgroup data including contents;
sql@ASM> alter diskgroup reco mount restircted;
sql@ASM> drop diskgroup reco including contents;

Let’s delete and recreate the grid disks. you can do it with dcli on all cells or one by the other.
The parameter SIZE it the important one to change the DG size! The RECO will get the remaining space if no SIZE parameter is specified.

root@db01# dcli -g my_cells -l root 'cellcli -e drop griddisk ALL PREFIX=DATA'
root@db01# dcli -g my_cells -l root 'cellcli -e drop griddisk ALL PREFIX=RECO'
root@db01# dcli -g my_cells -l root 'cellcli -e create griddisk all PREFIX=DATA,SIZE=3.1T'
root@db01# dcli -g my_cells -l root "cellcli -e create griddisk all PREFIX=RECO,CACHINGPOLICY='none'"

The warning for the Flash disks could be ignored, because the flash would be only used for caching. Now you can create your DGs:

sql@ASM> create diskgroup DATA 
normal redundancy disk 'o/192.168.10.*;192.168.10.*/DATA*' 
attribute 'AU_SIZE'='4M',
 'COMPATIBLE.ASM'='12.1.0.2.0',
 'COMPATIBLE.RDBMS'='11.2.0.4.0',
 'cell.smart_scan_capable'='TRUE',
 'appliance.mode'='true';
sql@ASM> create diskgroup RECO
normal redundancy disk 'o/192.168.10.*;192.168.10.*/RECO*' 
attribute 'AU_SIZE'='4M',
 'COMPATIBLE.ASM'='12.1.0.2.0',
 'COMPATIBLE.RDBMS'='11.2.0.4.0',
 'cell.smart_scan_capable'='TRUE',
 'appliance.mode'='true';

Please replace and expand the attributes regarding your needs.
On Exadatas prior X4-2, the infiniband interfaces was bonded as active-passive and the disk parameter must be adjusted accordingly ( ‘o/192.168.10.*/RECO*’ ). The syntax to specify the disks is really comfortable. After the creation the dg is only mounted on the node where it was created.Don’t forget to start you disk group.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s