VA7410 space issue

[Alberto Velez]

Hi all,

We have a new VA7410. It is configured on autoraid, but there is the same available space than space reservated for redundancy, which seems a high loss of space.

I've read that when the disk array is 50% full, it will begin to move data from Raid 1+0 to raid5 and new free space will be available. Is that true? How will be this space liberated, the already existing luns will be bigger or we'll have to create new LUNS? Does it have any problem working with LVM?

Thanks in advance :)

[Luk Vandenbussche]

Hi,

When you create more luns to that the capacity is used more then 50 % then the data will move from raid0/1 to raid5

LVM can perfectly handle this option, so don't worry.

[Ranjith_5]

Hi ,

Migration from Raid*0+1 to Raid*5 is an VA-internal process. Nothing to do with LVM. Only the lun redundancy level will be changed from Raid*0+1 to Raid*5. You are safe.

Regards,
Syam

[Alberto Velez]

Hi all,

Thanks for your fast responses. But my doubt is: how is this migration reflected to raid1+0 to raid5. I'll be able to create more luns, the existing luns will be bigger or I could store more date on the VA without decreasing the available space.

For example: I have a VA with 200GB, 100GB available and 100GB of redundancy. I create 4 LUNS of 50GB each. When the stored data exceed 100GB and the migration to raid5 is performed, will I have more available space to create more luns?

[Uwe Zessin]

The data will be migrated to RAID-5DP, not plain RAID-5. RAID-5DP redundancy requires less raw capacity than RAID1+0, because the overhead is smaller - so the storage array gains more free capacity, but the amount of user data that is visible though LUNs, of course, do not change.

[Alberto Velez]

So will I be able to create more LUNS when the data is migrated to RAID-5DP?

[Vincent Fleming]

When in AutoLUN mode, the array will allow you to create more than 100GB of LUNs (from your example) but not 200GB...

The exact amount of usable LUN space you will net will vary based on configuration.

If you post up your configuration, I can give you a relatively accurate estimate.

Regards,

Vince

[Alberto Velez]

Thanks for your response,

i have 2 redundancy groups of 333GB each, 135GB for redundancy,66 for active spare and 120 GB of available space.

the raid mode option is "autoraid", the data resilency as "normal", hot spare mode as "automatic", atutomatic include as "yes" and autoformat drive as "yes" as well. I'm using command view sdm 1.09.

I've created 4 luns of 50 GB each (2 in each redundancy group), so I only have 20GB available on each redundancy group.

I'm afraid I cannot find the autolun option.

[Vincent Fleming · Accepted solution]

What size disks are you using? 73GB?

I'm guessing each RG is 5 x 73GB drives.

In AutoRAID mode, you should be able to create approximately 185GB of LUNs in RG1, and 195GB of LUNs in RG2.

Don't forget that with AutoRAID, it will keep as much of your LUNs in RAID1 as it can manage, and convert the rest to RAID5DP (double-parity). Since you've allocated more than 50% of the RG's (100GB allocated in LUNs, ~ 190 GB available... it's more than 50%), then some of your data must be kept in RAID5DP. Don't forget that double parity means just that - there are 2 parity blocks per stripe instead of the usual 1.

Now, with 5 drives, RAID5DP is not overly space efficient - it requires 2 parity blocks. That leaves you with a 3:2 data:parity ratio.

At 3:2 (and the 73GB drives NET about 66GB), you net about 198GB usable space if it's ALL RAID5DP (ie: there is no RAID1.)

Now, deduct from that your spare(s) - 66GB each.

198 - 66 = 132GB actual usable space per RG.

In addition, there is some overhead by the AutoRAID - mapping tables take up a couple of GB, so you lose a little more space there.

So, your creating 2 50GB LUNs and having 20GB of free space per RG is about right.

The good news is that if you add drives, it changes the data:parity ratio - which means that you get to use a lot more of the additional drive space than what's in the array already. For example, by adding another 73GB disk to each RG, you'll actually get to use 66GB per 73GB drive (the actual capacity) - there will be NO ADDITIONAL parity space needed. It changes the data:parity ratio from 3:2 to 4:2.

Now, you also seem to have a question about how the VA moves the data from RAID1 to RAID5DP...

The RAID on the VA is at the data-block level, no at the physical disk level. What this means is that any individual data block on any LUN can be in either RAID1 or RAID5DP - and most of the time the LUNs are partially RAID1 and partially RAID5DP.

The movement of data from RAID1 to/from RAID5DP is done by the VA's controllers behind the scenes - meaning that the host is completely unaware that a change was made.

How much of your LUNs that are RAID1 vs. RAID5DP depends on how much free space you have in your RG's. You have 20GB free, so you will not have more than about 25% of the LUN space stored in RAID1. All the rest of the data would be in RAID5DP. This means that about 12GB of each of your 50GB LUNs will be RAID1, and the rest (38GB) will be in RAID5DP.

Does this help?

Regards,

Vince