Would like to move vsphere replication target to new datastore at remote site

[cypherx]

Hello,

We installed a new Tegile TE4100 disk array at our DR site.  We've already started moving some production workloads off our older EMC NX4 array.  Now what is the best way to move our vSphere replication data from the old array to the new array?

I can think that it would be possible to pause the replication, ssh to one of the ESXi hosts, do a cp command on the replicated data folder from the old filesystem to the new filesystem, then stop the replication (it will delete the data from the old), then create a new replication, point it to the new filesystem and data that was moved there and tell it to use it as an existing seed.

Or is there another, easier way (like a storage vmotion - but for vSphere replication data)?  If not, they really should put that feature in a future version of VR.

vSphere 6.0u2 , vSphere replication and SRM 6.1

[mvalkanov]

Hi,

Starting with VR 6.0, there is integration with SDRS.

If you put the two datastores in a datastore cluster and then invoke "Enter maintenance mode" on the old datastore, SDRS will tell VR to move the replica disks to the other datastores within the cluster. There won't be any additional checksums / seeding - the move functionality is provided by VR.

AFAIK this move functionality is not exposed in the VR UI of existing releases, but it is only an internal API being used by SDRS.

Regards,

Martin

[cypherx]

Well I'd like a little more control over what migrates and when.  It would be nice to right click a replication and have an option like storage vmotion right there.

Sounds like SDRS if I put one filesystem on maint mode, everything will move at once.  This is a phased approach.

I might be limited to the old, original way of pausing, copying data, stopping, then starting the replication pointed at the new location telling it to use that data as seeds.  A little painful, sure but its controlled.  I'd be open to other options.

[cypherx]

Let me ask this...

what happens when I click reconfigure replication on a VM and change the target datastore there?  Right now I have one that I did that on and it says "Configuring".  Is it smart enough to move the data locally?  Now it says OK after quite some time, I think it may have done it but I still see the data in both the old datastore AND the new, though more files in the new as the old datastore only has an hbrdisk.RDID, 3 hbrcfg files an hbrgrp file.  The new datastore contains an hbrdisk, the actual vm's name.vmdk file, and three hbrcfg files.

This arcticle in the vSphere Replication 6.1 documentation seems to elude to the fact that its as simple as clicking reconfigure replicaiton and choosing another datastore / location that is online and accessible.

VMware vSphere Replication 6.1 Documentation Center

It sure would be easier than a pause, mv command when ssh to an ESXi host, then stop, then create a new replication and point to the moved, seeded files.

[mvalkanov]

Hi,

Thanks for the doc pointer. I checked that the same topic exists in VR 6.0 documentation.

You can grep VRMS logs at the target site for "MoveCollection" this should confirm that a real move is being done as part of the reconfigure.

Regards,

Martin

[cypherx]

Where are these logs located?

I found /var/log/vmware/hbrsrv.log in the vSphere Replication Appliance at the target site, | grep MoveCollection returns nothing.  A tail of the log shows a lot of verbose hbrsrv Originator@6876 sub=SessionManager] hbr.replica.Task.GetInfo: authorized

[mvalkanov]

Hi,

hbrsrv.log is for the VR server.

/opt/vmware/hms/logs/hms.log is for VRMS.

Regards,

Martin

[cypherx]

I didn't see anything to those affect when grepping those logs at the remote site, however after I reconfigured some replication and verified the data was on the new filesystem, I did a test run in SRM and the machines came up without any issues at the remote site.  While they were temporally running in the test mode I made sure there was no remnant of them on the old filesystem and they continued to run (performance improved too over the new storage array).

The only thing I wish is that there was some insight to the progress when moving them.  I have 4 very large VM's currently that in vSphere Replication say "Configuring".  Unfortinately I have zero insight to how far the progress is.  The only thing I can go by is to use the datastore browser at the remote site and gauge the vmdk file sizes to see how much has moved.  A percentage bar like when a regular "sync" status would have gone a long way.  Hopefully in the next version, right?  Actually hopefully the next version will be HTML5 UI because modifying each of these takes a bit of time for the UI "barbershop poles" to populate all of the current (and new) details when changing the target storage.