Nested VMs

[Jim Taylor]

This thread belong in self-hosted, but for some reason I can't post there.
We host in a local data center running VMWare. We spun up a Windows Server and have Hyper-V installed. We have a SBC handling all our SIP traffic, and the Windows host sits on the same network as the SBC. We have a couple dozen Linux 3CX installations running on the host. My question: is anyone doing this type of setup to save on hosting costs, and if so, have you run into any issues? So far we have run very smooth, but rarely we'll experience some static or 3 second delay on the 3CX virtuals, but if our datacenter engineer moves the host machine to a different physical set of servers, the problem goes away. Anyone have any ideas?

 

[SweetAction]

Self Hosted is for your own account in AWS, Azure, GCP, etc.

Unsure how this setup saves you any $'s, but it's clearly something with the networking / hosting stack if a vMotion to another host fixes the issue. I would enable port monitoring, replicate the issue, and review the capture to find out more.

 

[SteveITS]

@Jim Taylor It's not uncommon for licenses to be miscategorized in the forum. https://www.3cx.com/community/threads/can-only-post-where-you-have-a-license-system.118970/

I didn't realize one can run Hyper-V underneath VMWare. I guess it's not new as the top search result is from 2013: https://4sysops.com/archives/how-to-run-hyper-v-under-vmware-workstation/. Why not just make 3CX its own VM? Is there a per-VM fee in this setup?

 

[Jim Taylor]

My datacenter host charges for cpu, memory, hard drive space, etc. as line items. CPU draw is usually very limited, so I can run a bunch of linux 3cx vm's and have fewer total cpu's on my hyper-v host, vs. 1-2 per 3cx instance. Same applies for RAM. I watch resources regularly and see how much RAM each instance actually uses, and it truly doesn't vary that much. I can issue fractional RAM from my hyper-v host and just aggregate the lot in a smaller amount than if I were to provision each vm independently. For larger instances (50+ users) I do have dedicated VM's on the datacenter platform itself, and when these very rare glitches do come up, it doesn't affect those instances, so clearly we know it's related to the nested vm's, just wasn't sure if anyone ran into this or tried this before.