Unsolved
This post is more than 5 years old
46 Posts
0
350
Optmizer + disk gaps
We've been migrating/killing some old raid5/6 luns and I now have some nicely swiss-cheesed disk groups with not only lun id gaps but lots of tiny gaps on the drives. The lun id gaps I can deal with by creating quantities of devices matching the id gaps... but I was wondering on dealing with the physical aspect.
I've never heard a good solution to it, seen some suggestions on creating smaller luns and making metas; but was wondering if optimizer would actually "clean up" my mess over time and give me a nice contiguous chunk of free space on the drives. Seems like it should technically be able to do this, we are on DMX4 + 5773 code so it can do a move to empty space rather than just swap, and moving everything into a targeted portion of the disk is what optimizer will do... For the hot luns, I'm sure it will move things nicely around but for our idle luns will it move those? Will optimizer move unused luns and make everything allocated nicely contiguous (and making any future raid1 configs easy), or will it leave those since just moving it is an expensive operation. I could probably try to coordinate hypers to disk gaps, match them to luns and mess with their optimizer priority level but that seems like a mind numbingly painful brain exercise.
RobertDudley
448 Posts
0
March 29th, 2010 05:00
Optimizer does a swap of hypers from one disk to another. Optimizer does not create a new hyper on unconfigured space and them swap or move your hyper to the unconfigured space. Optimizer will not do what you want.
InsaneGeek
46 Posts
0
March 29th, 2010 13:00
I'd agree that historically it couldn't but reading things it seems to imply it does.
the Symm Array Control CLI 7.1 doc" Under the sym optimizer migrate option
"When devices are moved to a new set of disks, this is called a migration. A migration
can be used to populate newly added disk drives, or to move devices between high
performance and high capacity disks. A migration can only be initiated by a user.
When devices are moved to a new specific location, this is called a relocation, and it
may be used to balance disk usage or to fill in holes left by deleting devices.
Relocations are performed by the Optimizer automatically, as needed. A relocation
cannot be initiated by a user.
...
When devices or device group devices are migrated to target disks or disk groups,
Optimizer looks for free space on the target disks for the migration. If the target disks
do not have any free space, you can specify to use unmapped and/or unmasked
devices. These options indicate that source devices can be migrated to space occupied
by unmapped and/or unmasked devices on the target disks. When this option is
specified, Optimizer will NOT look for free space on the target disks but will only
look for matching unmapped and/or unmasked devices."
So under that definition it seems like optimizer will fill in the blank spaces under "relocation" but there doesn't appear to be a "do that first" kind of option as it only does it as needed and can't be initiated by a user....
I tried to do a manal migration with a device with a target to the same disk group; but it failed complaining that "The migration request has been denied because all devices already reside on the target disks". So I can't just feed it a list of devices and let it decide where to put things, I would have to specify which physical disks.