Target Server Config regarding RDM Disks

[goldsmith]

Our production site uses a VMFS datastore for storing the OS drives of our VM's and RDM drives for storage volumes, when we configure a replication job the target server uses VMFS datastores only to create the target drives. Is there any plan to allow the Target machine to use RDM disks as target options, this would obviously require the target hosts to have the same LUN assignments and disks presented. I can't see this being too difficult as the underlying vmdk files act as a mapping file, so long as the target LUN assignments are consistent this should be achievable.

I really do hope this is in the roadmap as we would like to keep our replication site as close to the production site (with regards to config) as possible.

Many Thanks

Neil

[Gostev]

Are you talking about RDM disks in virtual mode?

[goldsmith]

Yes indeed, all disks are in virtual mode. It is the Target end that we would like to specify the RDM devices as target disks, also when will you be introducing the failback option?

[Gostev]

We do not support replicating to vRDM disks at the moment. The job will convert vRDM to regular VMDK on target. No plans to add full support in short term. Failback options, on the other hand is on high-priority features list.

[goldsmith]

Many Thanks Anton, sorry to bother you but have you ever discussed the target disk option and should I expect any news in the near(ish) future?

[Gostev]

Hi Neil, can you please clarify what you are asking about?

[goldsmith]

We would like to configure the target hosts to use RDM disks the same as the source, the traget host would be configured to use the same LUNs as the source hosts. We would then like the option to select the target LUN as a destination for the source disk during the Job wizard, or even if the Job wizard could identify the target LUN automatically rather than having the disks based in the VMFS datastore. I think this would be a good option as some applications are recommended to use RDM disks for performance, no doubt there is a good number of people with this kind of config. This will help with handling large data volumes, we have a number of volumes over 800GB and would like to avoid using extents on the datastore.

I hope this clears things up.

[Gostev]

This is possible today. You should exclude your RDM disks from processing in the job settings, but do not select Remove excluded disks from configuration checkbox. This will do exactly what you need.

[goldsmith]

Anton, fantastic! However, does this remove the drive from replication?

[Vitaliy S.]

Yes, If you exclude the drive in the job settings, it will be excluded from the replicated VM as well.

[goldsmith]

Vitaliy, thanks for that. As I supected, this negates the whole idea of replication as it would exclude the data volumes. I guess there is no choice other than to replicate the whole machine including large volumes to a VMFS datastore at the target end. Shame, would have been very nice indeed if we could map source drives to LUNs at the Target end. Or even replicate the machine without modification, this would allow the mapping files to target the disk accordingly.

Thanks for everyones input regarding this post.

[Gostev]

I am not sure I follow you, Neil. If you do exactly what I said, the when powered on replica VM will use those RDM disks that were previously connected to the source VM. Isn't it what you wanted?

[goldsmith]

OK, I think we might be getting our wires crossed here, the original question was if we could replicate source RDM disks using veeam to target RDM disks without them being created as regular volumes within the target datastore. The earlier posts said if I remove the drive from the replication job the drive will not get replicated (which makes sense), what I was looking to do was to select the target LUN when creating the replication job so that the source RDM volume is replicated to an RDM volume at the target end and not replicated as a regular volume and stored with the VMFS datastore.

Is this actually possible but I am missing something?

Thanks

Neil

[Gostev]

No, replication of RDM disks to RDM disks is not supported.

[goldsmith]

Many Thanks

[mdornfeld]

Neil / Anton,
I will most likely soon be doing replication and have just thought of the same issue with the vRDM's.

Neil, would you mind sharing how you ended up configuring your replication after this discussion took place?

Anton, I'm left wondering why what Neil and I do is so unique with vRDM's that this feature request isn't more common. From your perspective as somebody who works with a lot of different companies via Veeam, how do most companies handle VM's with large File, SQL, or Exchange drives on them?

Thanks guys,
Matt

[Gostev]

Vast majority are using VMDK for everything, because only this provides for complete storage virtualization. I like to tell this story, and probably will repeat myself, but one of our customers had an internal half-year research and testing project to choose between using physical RDM and virtual RDM for their large servers. This was way before they bought Veeam. Anyhow, they ended up going with VMDK, that was not even in the scope of their research initially because of bad perception of CIO

[mdornfeld]

Thank you Anton. Our vRDM strategy for large volumes may very well be outdated. I will have to reinvestigate this moving forward.

[mdornfeld]

With the implementation of some new storage, I will be migrating our vRDM's to VMDK's.

To do this, I have two options that should accomplish both the task of moving to the new storage and migrating the vRDM to VMDK simultaneously: Storage vMotion and Veeam Replication

I am considering using Storage vMotion for all of my smaller, lower I/O VM's

However for my very large VM's that currently incorportate vRDM's and high I/O, I'm considering using Veeam Replication. While this would incur a short outage while I permanantly flip over to the replicated machine on the new storage, I think I like having the security of not watching a storage vMotion take what I'm guessing would be hours upon hours and with Veeam Replication I'd have two solid copies of the VM and a solid backout plan.

Since I've never done this before, I wanted to see if there was any throughts on this migration plan from Veeam or other users (good, bad or indifferent).

Thank you all!

[mdornfeld]

If I did go the Veeam Replication route, with both storage pieces on the same SAN, as well as Veeam backup jobs for those machines being on the SAN, is the preferred method to still use the Veeam backup jobs for the initial seeding of the replication to keep the I/O stress off of the production disk of the original VM's? Thanks again!

[foggy]

Matt, your observations on converting vRDM's into VMDK's using VBR replication look very sensible. Also, a very good point on using replica seeding if you already have backups at your hands.