We have 2 separate DRS clusters: one for production virtual servers, second for development ones. Both clusters are on different networks, they share only SAN space. One of the suggestion was to join these clusters for better resource utilization. In my understanding joining will create additional layer in management - resource pool, plus prod and dev virtual machines still be on the same ESX server and use same resources.
I am looking for suggestions based on experience, best practices and etc.
Thank you in advance!
If you're planning to mix the PROD & DEV hosts within the same DRS cluster, you'll need to make sure all servers have access to all networks.
Regarding resource pools, This is an excellent way of seperating prod & dev vm's sharing the same physical hardware. You can set resource limits for your dev machines to ensure they don't take more than you require for prod.
Keeping the VM's on seperate LUNS would be a good idea to seperate prod from dev.
What is the benefit of mixing prod and dev, even if I separate VMs between resource pools?
I am in favor of separating because it will be easier to manage: no resource pools (except default one of course), dev and prod VMs are not on the same esx server, no need to manage resource pools and etc.Also in my opinion environment will be more stable.
What do you think?
Sounds like you already have the resources to seperate development from production so I would stick with that. Yes you could put them together and use software to limit development resources, but this would be like separating them with software. Development networks always seem to be the biggest problem with server sprawl. I would leave the setup as is.
The main benefit to running development and production together is that the redundancy overhead is less.
If you have two production hosts and two development hosts and want redundancy, you cannot use more than 50% of the total capacity of each of these clusters.
If you place all four hosts into the same cluster, you can push up utilisation to 75%, while still maintaining N+1 redundancy.
There is therefore a significant cost benefit in pooling as many resources together as possible.
The additional layer of management should be minimal. If you have different admins for each group and you need to setup rights for them to create VMs etc that will add a little extra work but not much. If you admin both there is no big deal.
The benefit is that idle capacity on one is used for the other. If there is any contention your resource pools can take care of it. Plus as others have mentioned you can share some of the failover redundancy.
My only concern is can you separate the networks. Do you have enough pNics in the sever to run seperate vswitches for prod and dev connected to different pNics. Thats going to stop any network load interference. Its not a must, just a nice to have.