Autogrowing thin volumes and LUNs

[HC_ITDEPT]

Hi all,

Just got my FAS2240 up and I'm trying to fire up my VMware environment.  My first task is setting up my ESXi cluster swap volumes.  These are probably the only volumes for which I will enable autogrow to avoid my vhosts freaking out if they need more swap space.  On my first cluster, I have set a volume as 50GB, autogrow up to 100GB at 1GB increments.  I then created the LUN as 48GB (actual size of physical memory), but I have no autogrow options in the LUN dialogs (I'm doing all of this in the OnCommand System Manager).  Everything is thinly provisioned and I'm using iSCSI.  How does the LUN know to grow into the volume, etc?  Am I supposed to set the LUN size to the full 100GB so that my vhosts see the max size and let the volume actually deal with the physical space consumption?

Sorry.  I'm sure this is asked elsewhere but I haven't found a satisfactory answer...

[aborzenkov]

Auto grow is for volumes, not for LUNs. NetApp won't grow LUN on its own.

20.11.2012, в 1:56, "Joey Prewett" <xdl-communities@communities.netapp.com<mailto:xdl-communities@communities.netapp.com>> написал(а):

<https://communities.netapp.com/index.jspa>

Autogrowing thin volumes and LUNs

created by Joey Prewett<https://communities.netapp.com/people/HC_ITDEPT> in Data ONTAP - View the full discussion<https://communities.netapp.com/message/94973#94973>

Hi all,

Just got my FAS2240 up and I'm trying to fire up my VMware environment. My first task is setting up my ESXi cluster swap volumes. These are probably the only volumes for which I will enable autogrow to avoid my vhosts freaking out if they need more swap space. On my first cluster, I have set a volume as 50GB, autogrow up to 100GB at 1GB increments. I then created the LUN as 48GB (actual size of physical memory), but I have no autogrow options in the LUN dialogs (I'm doing all of this in the OnCommand System Manager). Everything is thinly provisioned and I'm using iSCSI. How does the LUN know to grow into the volume, etc? Am I supposed to set the LUN size to the full 100GB so that my vhosts see the max size and let the volume actually deal with the physical space consumption?

Sorry. I'm sure this is asked elsewhere but I haven't found a satisfactory answer...

Reply to this message by replying to this email -or- go to the message on NetApp Community<https://communities.netapp.com/message/94973#94973>

Start a new discussion in Data ONTAP by email<mailto:discussions-community-products_and_solutions-data_ontap@communities.netapp.com> or at NetApp Community<https://communities.netapp.com/choose-container.jspa?contentType=1&containerType=14&container=2877>

[radek_kubka]

Hi,

May I suggest setting LUN space reservation to disabled (i.e. thin provisioning) & setting its size to the max value upfront?

Actually, if the volume is thin provisioned as well, I would probably set it to the max value as well, rather then using autogrow.

Regards,

Radek

[pascalduk]

Radek Kubka wrote:
Actually, if the volume is thin provisioned as well, I would probably set it to the max value as well, rather then using autogrow.

I would not set thin provisioned extremely larger than required.

WAFL scanners need to scan ALL blocks volume in case of e.g. snapshot deletions or deswizzling. Having more blocks than necessary will put unnecessary load on your system.

[radek_kubka]

WAFL scanners need to scan ALL blocks volume in case of e.g. snapshot deletions or deswizzling. 

Are you sure it also applies to blocks which have never been touched?

[HC_ITDEPT]

Thanks Radek.  When you say:

May I suggest setting LUN space reservation to disabled (i.e. thin provisioning) & setting its size to the max value upfront?

This is exactly what I was asking.  So what you're saying is, if both the volume and LUN are thin provisioned, I can make the initial LUN size LARGER than the initial volume size as long as my volume will autogrow to accommodate the LUN as needed, correct?  If I'm using this for VMware swap, will it get formatted with VMFS (or something similar) and actually chew up the whole LUN anyway?

Thanks for the help to all of you.

[radek_kubka]

In theory that may work - be careful though that autogrow reacts before the LUN gets full.

You need to select thin provisioning on the VMware side when formatting the datastore to preserve thin provisioning on the NetApp side.

[HC_ITDEPT]

Here's the part that scares me Radek:

In theory that may work

I don't have a test environment where I work, so I'm going to be doing this with production VMs.  Theories don't make me very comfortable.  What I'm trying to accomplish is the very thing mentioned in the TR-3749 manual, page 76.  If it is a best practice to use a "large thin provisioned LUN or a FlexVol with the autogrow feature enabled", then how am I supposed to accomplish this?  How do you guys do it?

[radek_kubka]

Personally I am a bit nervous whether vol autogrow will kick in on time if there is a massive, sudden influx of data into LUN - it should, but...

That's why I said I prefer larger volume, which is thinly provisioned, thus eliminating (or vastly reducing) the need for autogrow.

[aborzenkov]

I briefly tested and yes, you can make LUN larger, at least when space reservation is disabled.

simsim> df s32

Filesystem kbytes used avail capacity Mounted on

/vol/s32/ 19456 2200 17256 11% /vol/s32/

simsim> lun show /vol/s32/ttt.lun

/vol/s32/ttt.lun 10t (10995116277760) (r/w, online)