EMC SAN and Drive letters
Drive letters issue:
We have several backend exchange servers in an active passive cluster.
Currently each Exchange server has 2 Mailstores, running on an EMC clariion cx300i
We have 3 Database luns running raid 5, and 3 Log luns running raid1
These are on seperate disks within the SAN.
We have just added 2 additional shelves so we can add more mailstores, as well as RMSE for making clones.
My problem is this:
We would like to add an additional 3 mailstores per exchange server, as well as add another backend server with 5 mailstores. So 14 more mailstores in total.
I have already created the luns using raid 5 on 5 disks ( 300GB disks ) for the databases.
Do i need to add additional luns for the logs? I believe i do.. This would be neneficial anyway to get some extra spindles on there..
So, now that i have the luns, and have added the luns to the existing storage group, what is next?
I read something about a 64K offset. Not sure what to do there.
How do i partition and format the drive to make it usable?
What about running out of drive letters. As we continue to expand, we will soon run out of drive letters. What and how do i do about that?
Without drive lletters, how can i map drives to the server, so that the Exchange clusters will know where the data exists.
How do i get the exchange servers to see more than 1 data drive for the exchange servers?
right now each backend exchange server can see the san by 2 drive letters. 1 for data, and one for logs.
How can i set this up so i will have a bunch of data drives for additional mailstores.,
This all comes about because EMC best practices is to have a seperate lun for each mailstore and log DB.
So if we have 30 mailtores, and there are only 26 drive letters, minus A, C, D = 23 drive letters available..
Confused..
We have several backend exchange servers in an active passive cluster.
Currently each Exchange server has 2 Mailstores, running on an EMC clariion cx300i
We have 3 Database luns running raid 5, and 3 Log luns running raid1
These are on seperate disks within the SAN.
We have just added 2 additional shelves so we can add more mailstores, as well as RMSE for making clones.
My problem is this:
We would like to add an additional 3 mailstores per exchange server, as well as add another backend server with 5 mailstores. So 14 more mailstores in total.
I have already created the luns using raid 5 on 5 disks ( 300GB disks ) for the databases.
Do i need to add additional luns for the logs? I believe i do.. This would be neneficial anyway to get some extra spindles on there..
So, now that i have the luns, and have added the luns to the existing storage group, what is next?
I read something about a 64K offset. Not sure what to do there.
How do i partition and format the drive to make it usable?
What about running out of drive letters. As we continue to expand, we will soon run out of drive letters. What and how do i do about that?
Without drive lletters, how can i map drives to the server, so that the Exchange clusters will know where the data exists.
How do i get the exchange servers to see more than 1 data drive for the exchange servers?
right now each backend exchange server can see the san by 2 drive letters. 1 for data, and one for logs.
How can i set this up so i will have a bunch of data drives for additional mailstores.,
This all comes about because EMC best practices is to have a seperate lun for each mailstore and log DB.
So if we have 30 mailtores, and there are only 26 drive letters, minus A, C, D = 23 drive letters available..
Confused..
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
AnthonyP9618: - thanks for the article on mount points.
It's still not clear to me how this is accomplished in my situation. I am trying to create the mount point on our SAN so it's always available even if one of the servers goes down.
So basically i have this:
On the san i create a small 1GB lun that will hold the mount points.
I then go to one of the servers and use diskpart to set the offset to 64
I then use disk management to format the drive.
The next step is where im a bit fuzzy.
Do i create a drive letter at this point? or do i select "Mount in the following empty NTFS folder"
The latter doesn;t seem to work properly, and if i assign it a drive letter the other servers can;t see the drive letter..
What am i missing?
It's still not clear to me how this is accomplished in my situation. I am trying to create the mount point on our SAN so it's always available even if one of the servers goes down.
So basically i have this:
On the san i create a small 1GB lun that will hold the mount points.
I then go to one of the servers and use diskpart to set the offset to 64
I then use disk management to format the drive.
The next step is where im a bit fuzzy.
Do i create a drive letter at this point? or do i select "Mount in the following empty NTFS folder"
The latter doesn;t seem to work properly, and if i assign it a drive letter the other servers can;t see the drive letter..
What am i missing?
Can the other servers even see the drive?
On the CLARiiON, do you you have a storage group with all the hosts in the cluster added to it?
On the CLARiiON, do you you have a storage group with all the hosts in the cluster added to it?
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
So if you have 30 Stores, it would be 6 Data LUNs and 6 log LUNS. Then again seeing that you purchased 300 gig drives, that isn't the most efficient way to plan for exchange. Exchange should be planned with IOPS as the primary priorityand storage capacity as the second priority.
I would say make a RAID 5 Raid Group (5 disk) for each Storage group you have. Make 1 LUN for each, and put all the stores for that storage group on that LUN. So if you had 6 Storage Groups, you would need 30 spindles for data.
For logs, you need to keep them separate from your data. Data is random I/O and logs are sequential I/O. You want to try to keep random and sequential on separate physical spindles. For logs you could make a 2-4 disk RAID-10 (or RAID-1) raid group and carve a LUN for each Storage group.
I would recommend using smaller disks (146 or 72 gig) and not putting anything else on them except exchange until you gauge the performance using NaviSphere Analyzer. And only put 1 storage group on a RAID group. Don't try to double up.
FYI... How many users do you have and how big is the total of all your stores?
Also, the 64kb offset will save you a lot of headaches in the long run. RAID-5 random writes uncur a 4 I/O (2 writes, 2 reads) penalty for every 1 write. If you don't do the 64kb offset, you will double your penalty to 8 for some of your writes. Not doing the offset will strip about 20% of your performance away. You cannot do this once a LUN had been initialized in the OS. You need to do this before you format and out data on it. If you haven't done this, I suggest you make a new LUN, set the alignment and then move the stores over during a maintenance window.
Exchange is one of those things you have to get right the first time, or it can lead to a lot of problems later.
To make the drives usable, just go into computer manager --> storage --> disk management and initialize the LUNs you added to your exchange server's Storage Group (EMC CLARiiON storage group, not exchange)