Good day,
We have a small server stack in our operation, one of which is a dedicated blade server running Ubuntu 12.04LTS (we'll call this UBU). This server is our Cloudmin master running Cloudmin Pro 9.4.
Unfortunate for us, this is also the oldest server in the stack (as evidenced by running an EOL version of Ubuntu). In an effort to migrate away from deprecated technology, I set up a brand new blade server with CentOS 7 1810 with an installation of Cloudmin Pro 9.4 (we'll call this CCE7). PS - there are 7 other blade servers acting as KVM hosts on this network.
I followed the instructions as per: https://www.virtualmin.com/documentation/cloudmin/repl.
We now have UBU (master) replicating to CCE7 (Secondary). In theory, if UBU goes down, CCE7 should simply take over, should it not?
The concerns are that several critical services are not replicated, such as; DNS (Bind 9), DHCP (ISC DHCPd Version 4.1), postfix (not as essential but still on the radar).
All of our virtual machines are "hardcoded" during creation by Cloudmin to point DNS to UBU, which forwards external requests to our firewall. How do we go about accommodating this?
Is there a strategy to use whereby the aforementioned services are also migrated? Or do they need to be configured by an admin?
What about existing DNS configurations on VMs? Do they need to be changed manually?
Where we are headed with this is that we intend to simulate a "crash" of the master to allow the secondary to take over. The intent is to make CCE7 a standalone Cloudmin master so that we can retire UBU altogether.
Thank you in advance for any information you can provide.
--Chad
Comments
Submitted by JamieCameron on Mon, 04/01/2019 - 22:54 Comment #1
For DNS, couldn't you configure clients or your DNS registrar to use the IPs of both systems?
For DHCP, is the server being failed over a DHCP server, or does it use DHCP to get it's address?
Right now our Cloudmin Master (running Ubuntu 12.04) is both the DNS and DHCP server on the network (features that are installed during Cloudmin install). The DNS server has entries that support multiple domains; hardware, virtual, and external. DHCP only hands out addresses to physical systems being plugged into the network, not to VMs.
To the best of my knowledge, when a new VM is created in Cloudmin, it is assigned an IP from Host systems > IP Addresses > Common IPv4 pools. During creation, the VM is also assigned a DNS server (the IP of the Cloudmin Master) and searches the "virtual" domain. As I understand it, Cloudmin assigns static addresses to these VMs, so, do we need to manually change the DNS configuration on each VM to point to the Cloudmin secondary system (since it is running DNS as well)? Note, webmin configurations from the Cloudmin Master are not replicated during replication setup.
Should we be backing up configurations from the Cloudmin master using Webmin > Backup Configuration Files, selecting the services we need backed up, then restoring those backups on the Cloudmin secondary?
Alternatively, would we be wiser to virtualize our DNS servers? Separate them from the Cloudmin roles entirely? Or should the Cloudmin Master always run DNS and DHCP?
RE: DHCP - the server being failed over to (Cloudmin Secondary) has the DHCP service installed, but not configured. Its IP address is obtained from DHCP from the Cloudmin Master.
Thank you again.