MSA1000 running ESX3.01 - Need to Reconfigure from RAID ADG -> RAID10

[Alistair Nelson]

Hi

To configure the array in the first place I installed ACU on a Windows server. Now the MSA1000 is connected to ESX servers (DL385G2) so I can't run ACU (I understand the management agents don't provide this ability).

The HBA cards in the ESX servers are Qlogic dual port PCI-X.

I udnerstand the smartstart ACU also won't detect and manage the array through Qlogic.

So is my only option to bring in another Windows box, install a HBA card and run it up just for this reconfig? I believe this would be quicker than attempting it through the CLI interface, since I don't know CLI?

[Uwe Zessin]

I have just checked an old CLI guide and it seems to be fairly easy:

CLI> MIGRATE UNIT 0 RAID_LEVEL=5 STRIPE_SIZE=32

[Alistair Nelson]

Hi Uwe

Thanks for the comment. You really think it's only a couple of commands? You mention RAID 5, I actually want to go from RAID ADG to RAID 10.

So perhaps a command like that might work, anyone any experience?

[Uwe Zessin]

It was a direct cut&paste from the manual. Substitute "1" if you want striped mirrors (the SmartArray controller does not differentiate between 2-disk mirroring and striped mirrors).

The CLI has a pretty good online help, by the way.

I am not sure if applies to migrations, but you might want to change some priorites before you start:
SET GLOBALS EXPAND_PRIORITY=HIGH REBUILD_PRIORITY=HIGH

[Alistair Nelson]

Hi Uwe,

Thanks again. Maybe I will revisit the cmd line interface - tomorrow I'll setup the CLI console cable and have a look. If it really is this easy.

Are you suggesting the array can reconfigure from RAID ADG to RAID 10 without losing data (of course I will backup first anyway).

[Uwe Zessin]

Well, the purpose of the migration is to do it online and not loose data intentionally. Otherwise it would not make sense to me to implement such a function.

Of course I would do a backup, too. 'just in case'
It is a bit like an insurance: you pay so that nothing bad happens...

[Alistair Nelson]

Thanks. Also my current RAID ADG config has 1 hot spare. Will this automatically be included in the migration or would I need to remove the hot spare and add it to the logical drive somehow first?

Sorry for all the questions, first time, and your assistance has been highly appreciated...

[Uwe Zessin]

I am not familiar with your disk configuration, but if you switch to RAID10 you most likely need more space. Do you still have room in the disk drive array? If not, the MSA1000 will give you a message when you enter the MIGRATE command.

I have not tried it, but I do NOT expect that the MSA1000 uses the spare disk for that. A spare disk can be assigned to multiple disk drive arrays.

So you might need to add additional disk drives to the disk drive arrays before you can do the migration. You can release the disk drive from its spare function

CLI> DELETE SPARE UNIT=2 DISK109

... and add it to the disk drive array to gain space, but in order to figure out the exact commands, one has to know the configuration. That's no big deal once you have CLI access:

CLI> show disks
CLI> show units

See you tomorrow ;-)


> Sorry for all the questions, first time, and your assistance has been highly appreciated...

You're welcome. I do remember my first adventures with Fibre Channel RAID arrays like it was yesterday...

HP's ITRC forums has a similar 'reward' mechanism like VMware's VMTN, if you are familar with it.

It is described here:
http://forums1.itrc.hp.com/service/forums/helptips.do?#28

[Alistair Nelson]

Thanks Uwe. At the moment I have 7 disks in RAID6 + 1 hot spare (8 disks total). I need to convert to RAID1 now, assuming possible, and in 2 weeks I need to add 2 disks to the RAID1 array.

Hopefully I can delete the hot spare then run the migration, kicking off some backups now then I will give it a go.

Thanks
Al

(PS Yes I read about the points system when signing up to these forums)

[Alistair Nelson]

OK I have deleted the hot spare and am currently waiting for the existing logical unit to add the newly available disk.

This will probably not finish till tomorrow morning as the expand command configures the extra disk in the existing RAID6.

I tried to queue the migrate command (existing expand command is only 50% complete). Earlier today it complained that it was busy. But now it complains I don't have enough space?? :

CLI-1> migrate unit 0 raid_level=1 stripe_size=32
Current raid level : ADVANCED DATA GUARD (RAID 6)
Current stripe size : 16kB
New raid level : MIRROR FAULT TOLERANCE (RAID 1)
New stripe size : 32kB
There is not enough space available for this migration.
Unit migration failed.


How much space is required, my VM's are using only about 60% of the space in the existing RAID6 !?

Anyone any ideas? Your help much appreciated.

If I get really stuck I guess I can try the Windows interface again, I am hoping not to have to bring in and runup another machine just for this (hopefully) simple task!

Thanks

[Alistair Nelson]

I should also have reported the current disk config as requested by Uwe:

CLI-1> show disks
Disk List: (box,bay) (B:T:L) Size Speed Units
Enclosure 1: SCSI (0:06:00) PROLIANT 4L7I DB1.86
Disk101 (1,01) (0:00:00) 146.8GB 160 MB/s 0
Disk102 (1,02) (0:01:00) 146.8GB 160 MB/s 0
Disk103 (1,03) (0:02:00) 146.8GB 160 MB/s 0
Disk104 (1,04) (0:03:00) 146.8GB 160 MB/s 0
Disk105 (1,05) (0:04:00) 146.8GB 160 MB/s 0
Disk106 (1,06) (0:05:00) 146.8GB 160 MB/s 0
Disk107 (1,07) (0:08:00) 146.8GB 160 MB/s 0
Disk108 (1,08) (1:00:00) 146.8GB 160 MB/s 0

Notes:
The speed is the currently negotiated speed to the disk. This may
be less than the maximum speed supported by the device due to bus
faults, loss of signal integrity, etc.

CLI-1> show units

Unit 0:
In PDLA mode, Unit 0 is Lun 1; In VSA mode, Unit 0 is Lun 0.
Unit Identifier :
Device Identifier : 60060B00-00849540-A1F3DED7-51930003
Preferred Path : Controller 2 (other controller)
Cache Status : Enabled
Max Boot Partition: Disabled
Volume Status : VOLUME EXPANDING (45% complete) - by other controller
Parity Init Status: 0% complete
8 Data Disk(s) used by lun 0:
Disk101: Box 1, Bay 01, (B:T:L 0:00:00)
Disk102: Box 1, Bay 02, (B:T:L 0:01:00)
Disk103: Box 1, Bay 03, (B:T:L 0:02:00)
Disk104: Box 1, Bay 04, (B:T:L 0:03:00)
Disk105: Box 1, Bay 05, (B:T:L 0:04:00)
Disk106: Box 1, Bay 06, (B:T:L 0:05:00)
Disk107: Box 1, Bay 07, (B:T:L 0:08:00)
Disk108: Box 1, Bay 08, (B:T:L 1:00:00)
Spare Disk(s) used by lun 0:
No spare drive is designated.
Logical Volume Raid Level: ADVANCED DATA GUARD (RAID 6)
stripe_size=16kB
Logical Volume Capacity : 700,069MB

[Uwe Zessin]

I am afraid, that there won't be enough space even after the disk drive array expansion has finished:

The original was made from 7 x 146GB disk drives giving 1.022GB space. After adding the 8th disk drive, you should have a total capacity of 8x146GB = 1.168GB.

The logical volume has about 700GB. Twice the amount is needed for RAID10: about 1.400GB. So you are still short of about 2 disk drives. Sorry, but that is why I was asking for the numbers before telling you how to expand a disk drive array.

> How much space is required,
> my VM's are using only about 60% of the space

That does not matter, sorry. The MSA1000 serves pure data blocks and does not know what a file system is.


Oh by the way: be careful not to confuse the EXPAND command with the EXTEND command. The second one will increase the size of a logical disk, definitely not what you want now!

[Alistair Nelson]

Uwe,

Thanks for the information. Again Uwe this is very helpful. Of the existing 700GB array, I actually have about 270GB free in vmware filesystem. IE only 430GB is being utilised.

So I need to either add disks or

1) Shrink vmware filesystem (assuming possible)
2) Shrink logical volume (assuming possible)
?

[Uwe Zessin · Accepted solution]

You can't shrink a VMFS with data in-place.
Nor can you shrink a SmartArray logical disk.

You would have to backup your data (a great time to execise you backup plan, no?), then I would wipe the VMFS with the VI client and maybe zap the MBR to make sure nothing is left, delete the logical disk, recreate the logical disk, recreate the VMFS and restore the data.

Honestly, I am not sure if two more disk drives are sufficient, because I have not taken any controller meta-data into account (I simply don't have information how to do so, sorry).

And I don't remember if the MSA1000 firmware consistently displays "hardware GigaBytes" or if, for example, the logical disk size is show in "software GigaBytes". In that case, you would need a little more than 10 disk drives:

700.069*1.024*1.024*1.024*2/146.8 = 10.24

[Alistair Nelson]

This is now entering into a VMware rather than HP issue. I have image backups of all my VM's but this would mean I'd have to recreate all the VM's and VM disks.

Or see if I can just copy the VM disks and metadata? from the SAN to temporary storage.

We have 2 extra disks on the way soon anyway. That would be an option. But it sounds like 4 disks might be needed :(

[Uwe Zessin]

> Or see if I can just copy the VM disks and metadata?

Yes, absolutely! That's one of the great things with virtual machines.

I have done it the opposite way just a few days ago:
have a server with local storage and needed to replace the disks with larger ones. So I have connected a Fibre Channel storage array to the VMware ESX server, created a VMFS on it and moved all files over. Replaced the disks on the local storage, re-created the local VMFS and then copied all files back.

On another occasion I had moved two VMFSes between two attached storage arrays, just by "dd"ing the raw disk - that worked because the block sizes were the same and I "dd"ed in the right direction ;-)

[Alistair Nelson]

Uwe,

Thankyou, I backed up the data, recreated the LUN and VMFS, and all is ok now.