Start a Conversation

This post is more than 5 years old

Solved!

Go to Solution

2898

November 15th, 2010 09:00

LUN MIgration from Concatenated meta lun to Striped meta lun

Hello,

I have a meta lun with 2 components.

Comp 0 -> had 4 luns with 267 GB raid 1/0

Comp 1--_ had 2 lun with 402Gb raid 1/0

We want to get rid of the above lun and create a brand new lun with single component Comp0 and stripe the meta lun with 6 disk 1872GB and migrate the lun.

Is it really possible to do this.

131 Posts

November 15th, 2010 10:00

Yes.  Migrating from concatenated to striped LUNs is a normal practice.  There aren't any issues with this.

If I ever run into an emergency situation where I need to extend a LUN fast:  I concatenate additional space onto the existing LUN, then migrate to a striped LUN.

You can pretty much migrate a LUN (reserved LUNs,component LUNs not included) to any other unused LUN.  The underlying structure of the LUN doesn't really matter to Navisphere.  It will let you migrate all over the place, as long as the new LUN is the same size, or bigger.  The only real exception to this is a LUN used by a Celerra - Navi will let you migrate the LUN, but the Celerra won't like it.

542 Posts

November 15th, 2010 09:00

Yes.  as long as the new lun is the same size or bigger you can migrate it.  The underliying components of metaluns does not matter when doing aray based migrations.

So create your new lun the way you want it.  THen right click the old lun and select "Migrate".  in the window, select the new lun and a speed.  When it is finished,  you will need to expand it inside the OS if you migrate to a big lun.

258 Posts

November 16th, 2010 11:00

I have one more thing to add to this if we migrate the lun will there be restriping of data,  that do to the SQL index fragmentation ?

131 Posts

November 16th, 2010 11:00

Nothing.  LUN migrations don't just randomly place the data.  The host is what manages the files/index fragmentation and what files are saved to which blocks.  The fragmentation has nothing to do with the block-level stuff on the CLARiiON.

258 Posts

November 16th, 2010 13:00

Hello Kenn,

We need to bring down a server a repurpose a lun for this migration. So i have to wait till 9.00PM in order to do that. So i can start ASAP and swithc it to high in the morning or just start it at high and leave it there till it completes. How much performance imapct we will be seeing in the normal business hours for DB server.

542 Posts

November 16th, 2010 13:00

Why start it at 9PM.  just start it now and put it on High.  when you go home for the night, switch it to ASAP and check it in the morning when you come in.  You dont have to do anything for it once it starts unless you are migrating to a bigger lun.  THen you need to expand the disk at the OS

258 Posts

November 16th, 2010 13:00

Thank you soo much for the info Kenn

Our plan is to start @ ASAP at 9PM and decrease it to high around 7AM before the business starts. So probably i should be on the safer size, but one last question is i have the production lun combination of striped and concatenation not the migration is going to be done completly to striped lun so that doent make any difference right.

542 Posts

November 16th, 2010 13:00

Well if your looking to migrate fast,  ASAP is the best way but it can have a performance hit on your array.  I seen that it uses about 20% SP utilization on ASAP.  I can tell you from experieince that even on ASAP, 1.8T will not finish in 10 hours.  Why do you have a time limiit?  The source lun does not have to offline since the migration runs in the background. With a lun that big,  it might be better to start it at HIGH during normal working hours and then bump it up to ASAP when the IO is low.

131 Posts

November 16th, 2010 13:00

If any processing or backups go on on this array at night, don't do ASAP.

ASAP at FLARE 28 and below uses write cache.  Your write cache will fill up and very negatively affect performance on your entire array.

Use high - It will probably take a week to run.  There isn't any outage or downtime.  Your host won't even know that you're doing a LUN migration.  High on FLARE 28 also doesn't really take much of a performance hit.

542 Posts

November 16th, 2010 13:00

Actually the difference between HIGH and ASAP is pretty big.  Currently, i am migrating close to 100TB of data between 3 arrays and during these migrations, i have to consolidate down DAE.  I have already done hundreds of lun migrations and all of them have been on ASAP.  Now i know my environment pretty well and know when i can and can do them.  So ASAP is not bad in my opinion.  You just got to play it smart...

131 Posts

November 16th, 2010 13:00

1.8 TB in 10 hours - That's probably not going to happen.  You can always kick one off and wait 5 minutes and it will give you an estimate of when it will complete.

You can do a LUN migration while the data is accessible, it's not an outage.  Just bump the priority down during busienss hours.

What FLARE revision are you on?  If you are at 28 or below, don't use ASAP.  ASAP will fill up your write cache and make everything else slow, not just the LUN you need to migrate.

If you're on 29 and above - High isn't much slower than ASAP, but you'll get a lot less host impact.

I'd say high. ASAP isn't really good to use.

258 Posts

November 16th, 2010 13:00

Hello,

We are planning to start it late night around 9PM EST.

Currently the array is running flare 28. so i was guessing the ASAP would complete migration atleast 80% by tomorrow morning 9AM. so after that i will redue it to high, that what i wwas guessing. Help me if am wrong in selcting ASAP.

542 Posts

November 16th, 2010 13:00

Ya  just start it at ASAP. IT will give you a warning saying it could impact array performance but don't worry about that.   Is this migration from FC to FC?  at ASAP, it might be 3/4 the way done when you come in in the morning.

You could just leave it at ASAP and monitor the array.  if you see performance drops, then just throttle it down to HIGH.

As for the impact, this migration is done on the storage processor so any performance hits will be seen across your hosts attached to the SAN.  specifically luns that are owned by the SP that is controling the migration.

One thing i always do when i do these it to have the old and new lun controled by the same SP so only that SP will be affected.

258 Posts

November 16th, 2010 13:00

Hello,

Lun migration ASAP is preferred or High is preferred. Because we have 1.8TB lun and wanted to complete in 10 hours so. If some one can help on this it would be really great help.

258 Posts

November 16th, 2010 14:00

Yes Kenn it is between FC disk to FC disk so.

Thanks much for the sharing your experience. I would rather do as you suggested doing on SAME SP.

No Events found!

Top