Please start any new threads on our new site at https://forums.sqlteam.com. We've got lots of great SQL Server experts to answer whatever question you can come up with.

 All Forums
 SQL Server 2008 Forums
 High Availability (2008)
 Recommended Cluster Disk Setup

Author  Topic 

Gyto
Posting Yak Master

144 Posts

Posted - 2012-01-16 : 11:08:29
Hi there,

I will shortly be setting up a SQL cluster for the first time and I just wondered what the best disk setup would be?

I have seen the following recommended:

A: Quorum disk (1gb)
B: MSDTC disk (1gb)
C: Systems Databases (Data and Logs)
D: User Databases (Data and Logs)

This seems reasonable, but wouldn't it still be recommended to have the data and logs on separate drives? ...and if so would that apply for both the system and user databases? If you wanted them all on separate drives then that's going to be 6 drives which seems a bit overkill! Any thoughts?

Also, is 1gb enough for the Quorum and MSDTC disks?

Thanks,

Matt

dinakar
Master Smack Fu Yak Hacker

2507 Posts

Posted - 2012-01-16 : 12:39:24
At the minimum I would recommend:
1 LUN for TempDB Data
1 LUN for TempDB Logs
1 LUN for userDB Data
1 LUN for userDB Logs

system database data and logs can go on same LUN.

By the way, you dont need to assign drive letters to all disks.. SQL Server supports mount points, so create a root volume (with a drive letter) and create mount points for the other LUNs

Dinakar Nethi
************************
Life is short. Enjoy it.
************************
http://weblogs.sqlteam.com/dinakar/
Go to Top of Page

Kristen
Test

22859 Posts

Posted - 2012-01-16 : 13:51:38
... and backups on a separate drive
Go to Top of Page

Gyto
Posting Yak Master

144 Posts

Posted - 2012-01-17 : 05:16:44
Thanks Dinakar - that's the sort of thing I was thinking, although I wasn't aware of the LUN principle so thanks for that - fortunately I have server administrators to set this bit up for me, I just need to tell them what I want ;o)

The backups will taken to our central backup location, so I shouldn't need to worry too much about where the default location is.
Go to Top of Page

Kristen
Test

22859 Posts

Posted - 2012-01-17 : 06:38:19
I don't have experience of very large databases, but we backup to disk files (local-ish to the server) and then let the IT folk get those files onto Tape / offsite / etc.

Never been happy with direct-to-tape systems for a number of reasons (happy to elaborate if you need me to)
Go to Top of Page

Gyto
Posting Yak Master

144 Posts

Posted - 2012-01-30 : 08:07:47
Thanks Kristen - that's the way we will be doing it, backups to disk initially that are then archived off onto Tape.

One more additional question on this subject though...

I was under the impression that in a cluster with 3+ nodes, all nodes would share the same LUN for their UserDB Data files and the same LUN for their UserDB Log files, and that they would all be able to access and read/write to them at the same time. Someone else has questioned this saying they weren't sure if more than 1 node could access the same LUN simultaneously, so they might need separate data/log file LUNs for each node...?

I thought the whole idea was that they shared the exact same storage locations? Am I missing something here??

Thanks
Go to Top of Page

GilaMonster
Master Smack Fu Yak Hacker

4507 Posts

Posted - 2012-01-30 : 08:45:18
LUNs are only used by one cluster node at a time. Whichever node the LUN is active for is the only one that can access it. So if you have three separate instances (assume they're running on different nodes and are in different cluster groups), each needs its own drive for its files.

--
Gail Shaw
SQL Server MVP
Go to Top of Page

Gyto
Posting Yak Master

144 Posts

Posted - 2012-01-30 : 09:11:49
Ok thanks Gail, that's very helpful.
Go to Top of Page

Gyto
Posting Yak Master

144 Posts

Posted - 2012-02-01 : 11:44:08
Right, apologies for going over this again, but I just want to double check I've got the right idea before I go too far....

I am currently thinking at the following setup:

1 (LUN/Disk?) Quorum (1gb)
1 (LUN/Disk?) MSDTC (1gb)
1 LUN Node1 Data (User & System DBs)
1 LUN Node1 Logs (User & System DBs)
1 LUN Node2 Data (User & System DBs)
1 LUN Node2 Logs (User & System DBs)
1 LUN Node1 TempDB Data
1 LUN Node1 TempDB Logs
1 LUN Node2 TempDB Data
1 LUN Node2 TempDB Logs

Could someone possibly just answer the following questions:

1) Does each node have its own system databases or does the whole cluster share the same ones? (If they share they I will obviously only need 2 tempdb LUNs for the cluster rather than 4)
2) Aside from the tempdb, is it ok to put the other system databases on the same LUNs as the user db data/logs, or should I create separate LUNs for them?
3) Do the Quorum and MSDTC have to be on their own physical disks or can they be on LUNs?
4) Is 1Gb big enough for the Quorum and MSDTC LUNs/Disks?
5) Both the Quorum and MSDTC serve the whole cluster, so only one of each is required regardless of the number of nodes?
6) Can anyone see any other problems with this?

Thanks all!
Go to Top of Page

GilaMonster
Master Smack Fu Yak Hacker

4507 Posts

Posted - 2012-02-01 : 12:41:42
Each instance has its own system databases, and they are on shared storage. Remember that a LUN is owned by one node at a time

--
Gail Shaw
SQL Server MVP
Go to Top of Page

Gyto
Posting Yak Master

144 Posts

Posted - 2012-02-02 : 06:52:11
Thanks Gail, so can you confirm if this is correct:

In an active/passive environment, Each ACTIVE instance has its own system databases on the LUN(s) which it owns, then in the event of a failover the PASSIVE node simply takes ownership of these LUNs and effectively inherits the system databases contained on them (meaning the passive node does not have its own system databases to begin with?)

Therefore, all you'd need in a 2-node (1 instance per node) active/passive cluster is:

1 Disk Quorum
1 Disk MSDTC
1 LUN UserDb Data
1 LUN UserDb Logs
1 LUN SystemDB Data/Logs
1 LUN tempdb Data
1 LUN tempdb Logs
Go to Top of Page

dinakar
Master Smack Fu Yak Hacker

2507 Posts

Posted - 2012-02-03 : 00:19:21
active/passive is incorrect terminology. Its either a Single instance cluster or a multi-instance cluster. An instance can be active on only one node at a time. A node can have one or more instances active on it.. Each instance has its own set of disks that are owned by the node the instance is active on.

your list of LUNs are right though, for a single instance cluster. but "2-node (1 instance per node) active/passive cluster" is incorrect. There is only one instance and not per node.

Dinakar Nethi
************************
Life is short. Enjoy it.
************************
http://weblogs.sqlteam.com/dinakar/
Go to Top of Page

Sachin.Nand

2937 Posts

Posted - 2012-03-03 : 12:47:23
Any particular reason for selecting disk for MSDTC and Quorum ?

After Monday and Tuesday even the calendar says W T F ....
Go to Top of Page

deepaksahay09
Starting Member

3 Posts

Posted - 2012-03-10 : 05:10:34
your comment is right. I think this knowledge is very help full to me ,because it is interesting news.
unspammed
Go to Top of Page
   

- Advertisement -