vxvm migration

Wow! We have not done 'bottom-up' config in more that 10 years?

Not sure what your actual requirement is?

If you have an existing volume and want to migrate to new SAN storage, you can simply Initialize new lun and add to existing diskgroup, then mirror existing volume to new disk,
Wait for mirror to sync, then remove original plex.

See this TN for details: http://www.symantec.com/docs/TECH66011

If my understanding of your requirement is incorrect, please help us to get a better understanding.

Show us your existing config with :

vxdisk -o alldgs list

vxprint -ht

Then tell us which volume(s)/disk(s) need to be migrated.

Hi Marius,

You can not directly assign a subdisk to a logical volume, you will need to get subdisk assigned to a plex & then plex attached to logical volume.

Typically when you use "vxassist make" command to create a volume, vxassist will automatically create plex & subdisks depending on what "layout"you have chosen.

As you are saying, that there is already a "striping" (RAID 0) at SAN layer, I would recomend to create a mirrored logical volume in order to have redundancy. If you are not looking for redundancy, & one volume uses one single LUN, there is no benefit to create striping at vxvm layer as well as there is a striping already at SAN layer.

If you want to migrate the volume then to another SAN, you can get the LUNs from the target SAN presented to the host & add a mirror (plex) to the volume from target SAN. Once the mirroring is complete, you can disable the plex from original SAN, this will ensure your volume is served from target SAN. Once found working OK, you can delete the plex & subdisk from original SAN & take out the LUNs.

 

G

 

Hi G, What I wished to underline in case is that if I have a LV(of course mirrored),who has assigned a plex and this plex has assigned more than one subdisk,then in this case the I/o traffic is faster over the case with only one subdisk assigned to the plex.Even if the subdisk are created into a LAN within SAN(where is an stripping raid behind).My opinion is right or false? Migration. I was asked in context that the source LV is spanned across 2 or more subdisks. Sure the way you written is correct and is the way I perform.I am wondering only if in this case,the target LUN(who now is synchronized with the source LV by mirroring) is spread over 2 or mode subdisks as the source LV was. This is only for performance reason. Because the outcome of this migration is the LV spanned on one LUN,not on 2 or more subdisks pertaining to 2 or many LUNs. tnx so much, marius

Please show us your current config with 'vxprint -ht'.

When you add a mirror to the current volume, VxVM will attempt similar layout. 

You can change layout of the new plex when adding the mirror (e.g. concat, striped, etc) with vxassist command (layout = ....).

The current LUN layout on new storage array should already be configured for optimum performance (Raid-5 or Raid-0+1).

It is not a good idea to configure striped volume on LUNs that are already striped - unless you are 100% sure that they are not sharing underlying physical disks.

Hi, I saw you state "VxVM will attempt to create a similar structure" in migration. This means that if I have an LV on source, and this has one plex who has associated 2 subdisks,then after migration(mirror),the target LV will has one plex and the plex has associated 2 subdisks? tnx so much, marius

As per my post above:

You can change layout of the new plex when adding the mirror (e.g. concat, striped, etc) with vxassist command (layout = ....).

Hi Marianne, thanks.