Is it possible to combine two different SAN storage devices into a single drive in Windows Server?

Question

I’m a SQL DBA and my SAN admin said that they have to migrate few drives from HP 3PAR 7400 to EMC CX4. So SAN team wants to have a downtime on sql servers to migrate the drives.

What was am thinks was, is their any possibility to migrate the storage without bringing down the sql?

i.e I have D drive with 100GB of 3PAR is it possible to add EMC CX4 (100GB) to D drive and make it 200GB. later remove 100GB of 3PAR from D drive.

Please let me know is this possible?

Answer 1

What was am thinks was, is their any possibility to migrate the storage without bringing down the sql?

I will rephrase this question thus:

Is their any possibility to migrate ONLINE tables (clustered / heap) to another filegroup (=another files) ?

The answer is yes and it's very simple with Enterprise Edition. For example, assuming that current filegroup is CustomUserObjects (D:\MSSQL\Data\DataFile.mdf) and we have to move these tables to another filegroup NewCustomUserObjects (G:\MSSQL\Data\DataFile.mdf) then

[1] For clustered & nonclustered indexes could be used following approach:

CREATE UNIQUE CLUSTERED INDEX PK_TestTable_ID ON 
dbo.TestTable(ID)
WITH
(
    ONLINE = ON,
    DROP_EXISTING = ON
)
ON NewCustomFileGroup
GO

The same approach could be used also for NONLCLUSTERED indexes.

[2] For heap tables, one clustered index could be created (CREATE ... INDEX ... ONLINE = ON) followed by a DROP INDEX ... ON .....

More details here.

This approach has few limitations. The most important are:

[1] Some indexes (ex. spatial & xml indexes) could not be moved ONLINE.

[2] It will leave BLOBs ([n]varchar(max), varbinary(max), xml, etc.) values stored within old filegroup but there are some solutions.

[3] For FULL / BULK LOGGED databases, it will generate a lot of entries within database transaction log (ex. *.ldf) thereby affecting all functionalities based on Tx log: log shipping, database mirroring, AO Availability Groups, transaction replication, Change Data Capture, etc.

Answer 2

You need switchover to standby. If you don't have standbay server you need to create.

Your question is not correct. But I give you two answer (it is concept answers) 1. Teoretic yes it is possible. You need have a EMC powerpath and storage support. But if choose that way. You need to control segments for migration in other words it is IT risk:

Time for migration - in Global how much you have a time for migration to another storage.

I/O for disk - if you will be have a big I/O on the disk. Your service not will be available. and other moments....

2. Practice. Yes need downtime.

Answer 3

If you are using PowerPath as multipathing software, you can have no downtime migration from array to array using PowerPath Migration Enabler.

It establishes a replication pair, source (3par) and a target (cx4) volume. It does a bulk sync of the device on block level and replicates writes to the target volumes. After you decide to commit the change, it renames Powerpath pseudonames so your server is using the target volume as primary. Done!

docs: https://www.emc.com/collateral/TechnicalDocument/PowerPath-Migration-Enabler-5.7-User-Guide.pdf