Highlighted

Clustering with one physical and one virtual node with GCO

Dear Experts,

Need some expert opinion on design recommendations. 

Planning to have Enterprise Vault  in clustering configuration with VCS locally with two nodes, one physical node and one virtual node on a Hyper-v guest AND global failover with GCO to a physical server without local failover at remote site.

The OS is Windows 2008 R2 or 2012 R2.

Storage is HDS HUSVM.

Kindlty let me know if this is working solution.

1 Solution

Accepted Solutions
Accepted Solution!

Hello VFC is only available

Hello

 

  1. VFC is only available in HyperV 2012 R2 (if I recall correctly). The guests can be either 2008 or 2012, but please remember that VCS does not support EV on Windows 2012 R2.
  2. I would disable the site/tasks schedule during the time you perform the DR drill. You can then run a manual archive of a subset of mailboxes to prove that it works, but not the entire thing (unless its no that much data and you've got enough bandwidth.)
  3. Good point, we usually cluster SQL along with EV using VCS. That way you can create a VBS / Recovery Plan and fail over everything together. Its part of the USP that you don't need SQL admins or other teams involved when you're using VCS. You just click switch (or execute for the recovery plan) and it all fails over together.
  4. There used to be a defined method listed in the VVR admin guide describing how to use checkpoints and backup/restore to perform the initial sync. I've not tried that in years, we usually just start the sync and let it do its thing.

View solution in original post

17 Replies

I wouldn't do that... Plus..

I wouldn't do that... Plus.. Will it be supported in vcs?
Working for Quadrotech-it.com

personally, i dont care for

personally, i dont care for hyper-v but pretty sure it's supported. what are your concerns?

I was expecting a mixed

I was expecting a mixed response on this. My concern was your expert opinion if this should be a workable solution, Also failing over to a virtual machine from a standalone machine, if there are any caveats , if I should be concerned about anthing after the failback has occured.

Also after we fail back to remote site with GCO, I was wondering if turning off the archiving and only allowing the retrival would be a better approach so that a reverse resync wouldn't be neccessary.

Hi Rob, Thanks for reply. Why

Hi Rob,

Thanks for reply. Why do you think we shouldn't do it?

Changed the topic to SFHA

Changed the topic to SFHA Windows forums

Hi, This shouldn't be a

Hi,

 

This shouldn't be a problem but you have to configure VFC on HyperV 2012 (Parent) for mapping the storage from the SAN. Both physical and virtual should be running DMP.

 

Regarding your other questions about turning or archiving when you're in remote site... If you've performed a swtich to the remote site, that is not a DR test, you would have experienced a site failure or something else that was quite significant. So your exchange should have failed over with the rest of your apps. Also, if something bad really happened you probably wont even be able to replicate back to the primary because it would have been lost.

 

So you really would not want to or need to disable archiving. Depends on what happened.

Excellent points Riaan.. Is

Excellent points Riaan..

Is there any feature like VFC in 2008 R2? The Hyperv servers are all 2008 R2

About DR test, you nailed it... We would be perofmring the DR drills every 6 months and I want to make sure that the fail back is smooth... what would you suggest for DR drill?

Also two more questions - 

1. Shall we consider SQL replication(native) or VVR for the SQL server hosting the EV DBs. We would always want the stores/indexes and DBs to be in sync.

2. We are also planing to perform the initial seeding by backup restore ... after the restore is done, how can I sync the pairs? will it sync?

Accepted Solution!

Hello VFC is only available

Hello

 

  1. VFC is only available in HyperV 2012 R2 (if I recall correctly). The guests can be either 2008 or 2012, but please remember that VCS does not support EV on Windows 2012 R2.
  2. I would disable the site/tasks schedule during the time you perform the DR drill. You can then run a manual archive of a subset of mailboxes to prove that it works, but not the entire thing (unless its no that much data and you've got enough bandwidth.)
  3. Good point, we usually cluster SQL along with EV using VCS. That way you can create a VBS / Recovery Plan and fail over everything together. Its part of the USP that you don't need SQL admins or other teams involved when you're using VCS. You just click switch (or execute for the recovery plan) and it all fails over together.
  4. There used to be a defined method listed in the VVR admin guide describing how to use checkpoints and backup/restore to perform the initial sync. I've not tried that in years, we usually just start the sync and let it do its thing.

View solution in original post

Thanks a lot

Thanks a lot Riaan,

Understood all the points but the resync using backup/restore is still confusing.

""""When using checkpoints, you take backup of the data on the Primary and physically ship the backup media to the Secondary location, and restore the backup on the Secondary. When you start the backup, mark the starting point, by using the checkstart operation on the Primary. When you end the backup, mark the ending point by using the checkend operation on the Primary. While the backup and restore are going on, updates are written to the Replicator Log volume. To bring the Secondary data up-to-date, restore the block-level backup. After the restore is complete, start replication to the Secondary with checkpoint using the same checkpoint name that you had specified for the checkstart operation on the Primary.

The Secondary can be brought up-to-date only if the updates are still present in the Replicator Log volume. Using checkpoints is a multi-step process and therefore, needs to be done very carefully.""""""

When will the replicator log will be played? 

Also once the logs have been committed will the pair notify us about those logs being played?

Example—Synchronizing the

Example—Synchronizing the Secondary using block-level backup

This example explains how to synchronize the Secondary RVG hr_rvg on the Secondary host london with the Primary RVG on host seattle using block-level backup and checkpoint.

  • Start a Primary checkpoint on seattle:

    # vxrvg -g hrdg -c checkpt_presync checkstart hr_rvg

  • Perform a block-level backup of the data volumes in the Primary RVG.
  • End the Primary checkpoint when the backup is complete:

    # vxrvg -g hrdg checkend hr_rvg

  • Restore the backup to the Secondary data volumes.
  • Start replication using the checkpoint after the restore is complete:

    # vradmin -g hrdg -c checkpt_presync startrep hr_rvg london

Once you start the replication using the last command it will only replicate what was written since the start of the backup. It will then only have a small amount ot catch up on. Once it catches up its in sync.

Hi SymGuy-IT, Phyical to

Hi SymGuy-IT,

Phyical to Virtual GCO clusters are fully supported.  We don't limit the use of Physical and Virtual servers with the Window's SFW/SFW-HA products.

Thank you,

Wally

Thanks Wally for the

Thanks Wally for the confirmation.

Thanks Riaan for the time.

An update which I got from

An update which I got from support ... 

Kindly note: Such a configuration, may be possible, however it is not supported if any problems arise. Symantec recommends that the Operating System, architecture are exactly the same on both nodes, for the application to run seamlessly. If this is configured, it may work. However, we will not be able to troubleshoot and support any problems which may arise due to such a configuration. 
The query will remain same in case of Microsoft clusters.

Recommended : To use either both physical nodes OR both virtual nodes.

They are saying they dont support it .... Smiley Happy ???

Guys any ideas on this?

One of the major selling

One of the major selling points for VCS is that it can do P/V configurations..... Product Management should comment on this as this is in direct conflict to what the sales presentations state.

The engineer called back and

The engineer called back and reconfirmed that it is supported Smiley Happy Looks like h heard you ....

Still he is emphasizing on the fact that the failover can go wrong if the resources are not available on virtual node. 

""""I confirmed with Backline escalation and the physical + virtual node is possible. The failover will work. 
The only implication we may have is with performance, which would lie outside the scope of cluster. Performance issues may arise because of host machine resource allocations. 
However this is a supported scenario.""""

Haha, glad support can hear

Haha, glad support can hear me all the way from the desert.

 

Resource allocation must just be locked down from the hypervisor. Just ensure there is no sharing/pooling of resources. And get VFC up and running Smiley Happy

Cool. Thanks Riaan again .

Cool. Thanks Riaan again .