Hello, have a lot of (useless, most of them) posts on this issue here. Problem was always end up with the same problem: the storage. This is all solved now, managed to get a NetApp filer with 12 300GB 15k SAS disks, raidDP, snapshots, deduplication and so on. It is gorgeus.
Problem is I cannot get any hints from Cloudmin GPL as some of these features are PRO.
Now, my gear is:
-1 Cisco SG30028 L2/L3 switch, working in L2 now;
-1 IBM X3650 M2, 24GB DDR3, dual Quad Xeon, 6 10k SAS disks;
-2 HP DL385 G5, 32GB DDR2, dual Quad Opteron, 2 10k SAS disks;
-1 NetApp FAS2020;
I have 2 connections from 2 ISP (no BGP, 3IPs for each one). ns1.xxx.ro (the IBM) on one provider and ns2.xxx.ro (a virtual machine for now) on the other.
Now I am thinking something like this:
-use NFS on the filer, CentOS 6.4 as the OS, KVM for virtualization;
-have to aggregate some nics to get, failover&speed in the internal LAN;
-use the IBM as the machine running all the services and domains with Virtualmin Pro (4 of the drives are a raid5 /home partition, the other 2 in raid1 = the OS);
-use the IBM as the Cloudmin controller, and the HPs as the hosts for a few virtual machines, their /home partitions on the filer; also in case of IBM fail I should be able to quickly (automatically?) to move the domains on the HPs (on the physical host or in a VM, doesn't matter;
-how can I achieve some Internet connection redundancy in this setup? If one goes down the other connection to take over (DNS involved);
-are there any issues now between the Intel and AMD CPUs regarding the virtual machines? A few years ago all hypervisors had problems migrating a vm from one host to another with a different CPU. Anyway can I keep the IBM out of the VM stuff, just to controll it?
-what should I do (how to setup things to help me) in the case of a controller meltdown ie IBM fail?
-recommended DNS setup?
What would be your thoughts on a "Perfect Cloudmin&Virtualmin Pro Setup with CentOS 6.x"? A "How to Forge" this step-by-step guide would be welcome :)
Howdy,
Yeah, redundancy is a complex topic, and the primary documentation we have on that is for Cloudmin Pro, which has some features built into it to handle that.
It's absolutely possible to achieve a redundant setup outside of Cloudmin Pro, but it's a hugely complex topic, and varies wildly depending on hardware -- and due to that, we don't have any specific instructions on how to do that.
The following link is a Support Ticket containing some details on how setting up redundancy works with Cloudmin Pro (as well as a mention of a common way of setting up redundancy without Cloudmin Pro):
https://www.virtualmin.com/node/28962
Hello - not outside your products I am not asking this, of course. What I ment was that I would like to use the IBM as a physical machine for Virtualmin PRO, and as a controller or something maybe for Cloudmin. But not to host virtual machines; those should be hosted on the 2 HPs and on the filer.
Let's take this step by step (BTW, it's probably why ppl keep asking this, you can't relate to the documentation, wich explains a few options, that you can't test; it's not a "how to"/"hands on"). But the main problem is that you don't explain certain things, for example what can you do for the Virtualmin domains in a Cloudmin setup.
Exactly which one is true:
-the domains hosted with Virtualmin PRO should reside in a physical machine;
-the domains hosted with Virtualmin PRO should reside in a virtual machine;
-it doesn't matter! Cloudmin will rule them all!
-this just can't be done, our software doesn't do that.
for automated failover?
Just go with me and you will see why I am asking despite the link.
Don't take the name of root in vain...Howdy,
I really was attempting to tackle the questions you're asking with the link I shared :-)
What I may end up doing is taking the details from that Support Ticket, and using that as a basis for some additional documentation to try to cover some of the concerns you shared.
Virtualmin Pro isn't a key component in this though... whether you're using Virtualmin Pro or GPL doesn't matter.
If you want automated failover, and you want a Virtualmin product to do it for you, then you'd need Cloudmin Pro.
Cloudmin Pro would reside on your controller machine (it doesn't even need Virtualmin), and monitor the other systems. When there's a problem, it can help in one of these two ways, depending on your setup --
If you have a SAN or other shared storage available to you, and your Virtualmin servers are VPS's running off that SAN, you could use this failover system here:
https://www.virtualmin.com/documentation/cloudmin/vm/failover
The advantage of it is that when the backup server comes online, it's using the disk image stored on the VPS server, so it would be in the exact state the previous server was in, with all the data that was there.
As soon as the new system can boot, it can be serving requests exactly as the previous server had been.
Alternatively, if you aren't using VPS's and shared storage, Cloudmin can be configured to monitor your primary server, and if it goes down, can switch the DNS records to point to a backup server. That's described here:
https://www.virtualmin.com/documentation/cloudmin/vm/roundrobin
If you use that second system, you could use the following in order to replicate your domains and settings from your primary server to the backup server:
https://www.virtualmin.com/documentation/cloudmin/vm/vsync
You have the option to specify how frequently data is sync'd between the primary and backup servers.
So the answer to your question is that it depends :-)
Your Virtualmin systems can reside in a VPS if you want to do the first type of failover, or a physical system or VPS if you want to do the second type of failover.
-Eric