2010-07-08 12:00 PM
Hello, I am soon going to be re-creatng an aggregate that will be used for our vmware environment, and I just wanted to ask for some tips and best practices that I should implement.
So here is a little bit of info about my filer.
Its a Fas3140 filer, running ontap 188.8.131.52p3
Currently I am in the process of moving volumes from this Aggregate of fiber channel disks, to an SATA Aggregate, they are all cifs volumes.
If i do aggr status <aggrname> -v I get the following;
Aggr State Status Options
cifs online raid_dp, aggr root, diskroot, nosnap=off,
redirect raidtype=raid_dp, raidsize=16,
Volumes: vol0, finservacct, it, marketing, pmdata,
citrix_streams, users, gms_data, data, gcs
Plex /cifs/plex0: online, normal, active
RAID group /cifs/plex0/rg0: normal
RAID group /cifs/plex0/rg1: normal
I know that this aggregate is using raid_dp and has 16 disks.
if I do df -Ah cifs:
Aggregate total used avail capacity
cifs 4087GB 3125GB 962GB 76%
cifs/.snapshot 215GB 61GB 153GB 29%
I see that its also taking up some space by doing aggr snapshots, which I will disable the schedule and purge those snapshots later on.
So I have about 4293GB if I take back the snapshot space.
We dont have an NFS license, we have a VMFS, so the new volumes on that aggregate will be vmfs. Is there a good way to split up that 4293GB of disk space into 2 or 4 volumes, which then on will contain a lun presented to esx?
2010-07-08 02:19 PM
What size disks do you have and how many available for a new aggregate? I recommend aggregates as large as possible. For example:
300GB FC disks:
3x RAID groups of 21 disks each.
450GB FC disks:
2x RAID groups of 22 disks each.
1TB SATA disks:
2x RAID groups of 12 disks each (only 23 will fit).
The performance from aggregates like this is astounding. With aggregates like this, I am never worried about I/O cross-contamination.
Professional Services Consultant
Virtualization and Consolidation
NCIE-SAN, NCIE-B&R, SCSN-E, VCP
919.250.0052 Direct Phone
2010-07-08 10:17 PM
Is there a good way to split up that 4293GB of disk space into 2 or 4 volumes, which then on will contain a lun presented to esx?
I am not sure I really understand a question. You just creates FlexVols on this aggregate (like you already did) and then create LUNs inside these FlexVols. Or could you explain in more details what you intend to do?
2010-07-09 06:14 AM
Ok, so what if I will add 5 more disks to this aggr, is it a good amount, do I need more or less is what I am trying to find out for performance.
Also, once the aggregate is ready to be split up into volumes, whats the most optimal way to divide it up?