| Issue 58: | FLOW of OHSM | |
| 1 person starred this issue and may be notified of changes. | Back to list |
I want to know the flow of ohsm once he has configured ohsm on his system As per i know -> the user will boot the system -> he will activate the lvm say /dev/ohsm_vg/lvol0 ( command -a y ohsm_sb) -> he will mount the lvm on any folder say "/mnt" -> he will start his work on the mounted file system -> after some time he may feel that he needs to activate ohsm -> he uses the command "ohsm -e -a allocate1.xml -r relocate1.xml -i tier_info.xml -m /mnt -d /dev/ohsm_sb/lvol0 -> he uses the mounted file system normally
Feb 1, 2009
Project Member
#1
sandeepksinha
Feb 1, 2009
At the very starting phase i have 3 primary ioctl's that will work. ./ohsm -f /dev/ohsm_vg/lvol0 This ioctl will give information about all the underlying devices.Thata all. ./ohsm -u /dev/ohsm_vg/lvol0 This will invoke the dm_ioctl for mapping of the devices.Because once the user knows from the initial ioctl that these are the devices below and if he is ready to use this LVM only the he can call the mapper function directly. ./ohsm -l /dev/sda7 /dev/sda6 This will create a logical volume over this devices and will call the mapper function once LVM has been created.This will be used if the user wishes to have a seperate LVM for these particular devices.
Feb 1, 2009
Looks Good NEO. Good work for providing such beautiful IOCTL.
Feb 1, 2009
I dont think we need to call the ioctls for making the mapper_table with the options " -u " and " -l " because the mapper table is needed only when the user wants to use OHSM what if the user does not enable ohsm after that : the ioctl is wasted moreover while enabling if the user does not provide me with logical volume name the enable option will not be able to call the DM ioctl The lvol name can be hardcoded only when we force the user to make the logical volume using my scripts and commands
Feb 1, 2009
As far as we have deciced ioctls wiil be called only in three options -> ohsm enable " -e " two ioctles one to dm, another to /dev/ohsm -> ohsm disable " -s "(not deciced) one ioctl to /dev/ohsm -> ohsm relocate " -o " one ioctl to /dev/ohsm user will have to give the logical volume name whenever he has to enable the OHSM
Labels:
-Priority-Medium Priority-Critical
Feb 1, 2009
The following is the complete flow of OHSM: -> the user will download the patch from the website -> untar the patch -> applies the patch on a preinstalled kernel -> applies the patch on a fresh unpacked kernel -> in any case he will compile the kernel with our patch again -> make modules_install for dm and ext2 -> goes to the folder of admin -> compiles the code for admin -> compiles the user space program -> reboots the system -> make a logical volume say "/dev/vg/lvol0" -> make an ext2 FS on it -> mounts the FS on any folder say "/mnt" -> goes to the admin folder -> inserts the admin module -> mknods with the major and minor minor number obtained -> edits the sample policy files according to his needs -> runs command for enabling the ohsm with options -e -> full command is : ohsm -e -a allocation policy file name -o relocation policy file name -i tier device policy file name -m mountpoint -d device name of LVM -> while disabling the ohsm the user will use ohsm -s for stopping -> for relocation using the ohsm the user will enter ohsm -r 1 3 5 where 1 3 5 are the rule numbers
Feb 1, 2009
Well mark the IOCTL for listing of allocation and relocation rules for OHSM 1.0.1 meaning it should be there for BITS.
Feb 1, 2009
Also see issue 59 for implementation details
Feb 2, 2009
Hi Rishi, What happened to the doc that I asked you to write "How OHSM Works" ?? I should be just like a nice OHSM Admin guide, with examples and sample XML files. If that document is completed then upload it and close this issue!
Feb 2, 2009
The that document will be released soon....
Labels:
-Type-Defect Type-Task
Feb 21, 2009
Rishi where is the document ??? Its not yet released what are u doing ....
Feb 21, 2009
That document will be released with OHSM 1.0.1......till then stay tuned ;-)
Feb 26, 2009
Closing the Issue
Status:
Done
|