I have a customer who has two vCenter Servers, version 6.0, running in Enhanced Linked Mode. Due to organizational changes, we need to separate the two vCenter Servers from each other, that is, break the Enhanced Linked Mode.
We would like to keep both vCenter Servers running as is due to integration with other products and not have to kill one of them and install a new instance.
Is there any (supported) way to split the servers in Enhanced Linked Mode?
@Adam: Can they back up the vCenter database and reuse that for the new install?
@Matts: Are you using Windows based vCenter servers or the VCSA? Version 6.0 or 6.5? Embedded or External PSCs?
Depending on your configuration, it may be *possible* to split. If you could provide a bit more information about the setup (Windows or VCSA, External or Embedded PSC, etc) that would help us determine if there's an option.
Even if he is currently not using external PSCs he could
a) create a new one and join it to the existing SSO domain and site
b) repoint the second vCenter server to the new external PSC (as per How to repoint vCenter Server 6.x between External PSC within a site (2113917) | VMware KB if coming from another external PSC, or using Reconfigure a Standalone vCenter Server with an Embedded Platform Services Controller to a vCenter S... when coming from an embedded PSC)
c) disable mutual replication between the new PSC and all other PSCs (as per Determining replication agreements and status with the Platform Services Controller 6.X (2127057) | ... )
This would effectively isolate the second vCenter server from the first one and make it use it's own PSC.
Right, that's another method. Hence, I wanted to understand the configuration before providing further guidance. In both scenarios, to be safe, you should "network isolate" the PSCs from talking to one another and then on each respective env, perform a standard "unregister" operation (using cmsso-util, no need to manually mess w/replication agreements as unregister operation will handle that for you) of the other PSC/VC nodes, this would be like a standard decomission operation which we fully support, except we're not actually decomissioning the systems but keeping them running as independent instances.
I've pinged a few folks internally asking about this, but my gut feel is that this may not have been an explicit scenario that we've tested internally. So though this will absolutely work (just leverages supported tools) and I know of others who have done this as well for this very scenario, it may require customer to file an RPQ if you're looking for official support ... I'll see what I hear back
Very interested in this topic as I have a customer requesting this right now.
My thoughts was similar to lamw's description of firewalling the environments from each other and "decommissioning" the opposite sides PSCs.
Theoretically this shouldnt cause any issues since, afaik, every PSC replicated holds all information, but when asking the support about this, the reply has been that it is an unsupported action.
Just wondering if its unsupported due to potential problems arising or if its just not tested.
and welcome to the forums!
As William Lam pointed out this procedure works technically without issues arising, but it is unsupported by VMware, because it has not been tested as part of the product's quality assurance (QA).
If you want this to be supported for a production environment then you can try to do an RPQ (Request for Product Qualification) with VMware to have them test and support this for your specific environment. I'm not sure about the process to achieve this - if you are a TAM customer I suggest that you reach out to your TAM.
Thanks! Yeah I figure that is the only way to go about it and that process is too long for our current project but will be something the customer will look into when done. Will just keep the environments together for now.
Was kind of hoping against hope that William would come back with some good news after checking internally.
In talking to PM/Engineering, this particular scenario is not something we've explicitly tested, so we wouldn't be able to officially support it. This is definitely something we would like to support and enhanced in the future, so PM/Engineering is aware of the use case. In mean time, for folks that are interested in the step by step instructions, I will be publishing a blog article later this week.
Just one information related to the supportability of this operation.
I think nothing changed so it's still not an officially supported method, is this correct?
Exactly what is not supported? Just the operation of "splitting" the vcenter/spc?
At the end of the operation the 2 resulting domain will be supported or they'll be unsupported as they are the result of this not officially supported operation?