Skip to main content

Hey legends,

 

Hope you’re all doing great! Always appreciate the insights from this community. Got a quick one for you.



We got Commvault VM replication running for a SQL VM with an RPO of 4 hours—kept it that way since it’s crash-consistent, and we already got SQL agent backups (t-logs every 15 mins + daily fulls) covering us.

 

Need to run a bubble test, and Commvault suggested a test failover + SQL restore, but the failover VM spins up with the same IP as prod on an isolated network. cant move it to routable network as we can’t have it talking to AD/DNS since there’s no firewall or ACLs to block it, so we’re thinking of using a proxy with dual NICs: 1 on isolated network and one on MA/CS network
 

This should let us restore SQL data to the test VM without exposing it to prod. But would the failover VM still route through CS/MA and reach AD/DNS? We only want restore traffic going through, nothing else.

Can we set up a one-way tunnel in Commvault to allow restores but block all other traffic?

 

Not using SQL replication due to licensing constraints and not doing continuous replication for the same reason.

Anyone done somethin similar? Open to ideas.

 

Anyone come across this before?

I'm currently working on this with Commvault support, but it could to be outside their typical scope. Has anyone dealt with a similar situation or have any insights to share? 


This solution sounds ripe with possibilities of things going wrong.  I have witnessed bubble tests going wrong when small things get overlooked.  Personally I would ditch the VM replication and go with CommServe Live Sync for its flexibility and supportability.

Thanks,
Scott
 


Thanks for your response ​@Scott Moseman , appreciate the insights! Totally get where you're coming from on bubble tests getting messy.

In this case, the customer is leaning towards VM replication with SQL backups for point-in-time restores, it seems like the simpler path without introducing extra licensing challenges. So chose the proxy path but want to ensure there's no way for the bubble VM to talk back through the CommServe.  Was thinking of setting up a one-way tunnel to control the communication flow, but wanted to check if that’s the right approach. Any thoughts on whether this setup could cause any unexpected behavior, or if there’s a better way to lock it down?


Oh, I misinterpreted the request.  My mind was thinking about a CS failover, not just a SQL VM restore.  I still consider a bubble test to be messy.  An out-of-place SQL restore will not satisfy them?

Thanks,
Scott
 


Reply