Saturday, September 17, 2011

Creating a block device for ASM from a file using iscsi target in RHEL 6

When we want to use a disk for ASM and do not have any unpartitioned space on system, we can create a file on disk and use it as disk with steps below. For this, we are going to create a file in /vdisks and map it as block device /dev/asmdiskA. 

1. Create directory /vdisks
# mkdir /vdisks

2. To share the files in this directory using iscsi target, we need to set SELinux context tgtd_var_lib_t to it.
# semanage fcontext -a -t tgtd_var_lib_t "/vdisks(/.*)?"
# restorecon -R /vdisks
# ls -ldZ /vdisks
drwxr-xr-x. root root system_u:object_r:tgtd_var_lib_t:s0 /vdisks

3. Create a file asmdisk1.img of 2GB. SELinux context of the file will be automatically set to tgtd_var_lib_t.
# dd if=/dev/zero of=/vdisks/asmdisk1.img bs=1024 count=2097152
2097152+0 records in
2097152+0 records out
2147483648 bytes (2.1 GB) copied, 27.8373 s, 77.1 MB/s

# ls -lhZ /vdisks
-rw-r--r--. root root unconfined_u:object_r:tgtd_var_lib_t:s0 asmdisk1.img

4. Configure iscsi target by writing below lines at the end of file /etc/tgt/targets.conf to share the file using iscsi target.
<backing-store /vdisks/asmdisk1.img>
vendor_id MYVENDOR
scsi_id 456789

5. Start iscsi target, make it to be started automatically while booting the OS.
# service tgtd start
Starting SCSI target daemon: [ OK ]
# chkconfig tgtd on

6. Start iscsi initiator and discover the targets. Here target is on same machine, so we use loopback address (
# service iscsi start
Starting iscsi: [ OK ]
# iscsiadm --mode discovery --portal --type sendtargets
Starting iscsid: [ OK ],1

7. Once target is discovered, we can login to target using command below.
# iscsiadm --mode discovery --type sendtargets --portal -l,1
Logging in to [iface: default, target:, portal:,3260]
Login to [iface: default, target:, portal:,3260] successful.

After this, a new disk is visible as under fdisk -l output.

# fdisk -l
Disk /dev/sda: 500.1 GB, 500107862016 bytes
..trimmed for clarity..
Disk /dev/sdc: 2147 MB, 2147483648 bytes
..trimmed for clarity..
Disk /dev/sdc doesn't contain a valid partition table

8. Now we have the disk available, but it will not available after reboot (even if we execute chkconfig iscsi on). The reason is, target will be started later initiator. So, we have to change the order of starting services while booting the service.
# cd /etc/rc5.d/
# ls *tgtd*
# ls *iscsi*
S07iscsid S13iscsi

For this we should edit the scripts /etc/init.d/iscsid and /etc/init.d/iscsi like below.
Change this line
# chkconfig: 345 7 89
# chkconfig: 345 40 89

Change this line
# chkconfig: 345 13 89
# chkconfig: 345 41 89

When we execute the following commands it recreates symbolic links in rc*.d directories with desired numbering.
# chkconfig iscsid off
# chkconfig iscsid on
# chkconfig iscsi off
# chkconfig iscsi on
# cd /etc/rc5.d/
# ls *tgtd*
# ls *iscsi*
S40iscsid S41iscsi

9. The block device attached now as /dev/sdc, but we want this name to be persistent with a custom name /dev/asmdiskA. And also this should be owned by user oracle, and group oinstall. So, we have to configure udev rules accordingly based on scsi id of the device.
We can find scsi id using the command
# scsi_id --whitelisted --replace-whitespace --device=/dev/sdc

The udev rule we write here is to map the device as /dev/asmdiskA with owner oracle, group oinstall and 660 permissions.
Create a file /etc/udev/rules.d/55-asmdisk.rules with the line below . The entry for RESULT should be what we got output as scsi_id command output above.
KERNEL=="sd*", SUBSYSTEM=="block", PROGRAM="/sbin/scsi_id --whitelisted --replace-whitespace /dev/$name", RESULT=="1456789", NAME="asmdiskA", OWNER="oracle", GROUP="oinstall", MODE="0660"

Once we restart iscsi service, /dev/asmdiskA will be ready.
# service iscsi restart
Stopping iscsi: [ OK ]
Starting iscsi: [ OK ]
# ls -l /dev/asmdiskA
brw-rw----. 1 oracle oinstall 8, 32 Sep 16 11:23 /dev/asmdiskA

After completing all the steps above, we can just restart the system just to ensure it is available after reboot also.

No comments:

Post a Comment