[PVE-User] Proxmox VE Clustering Usage Scenarios

[Theron T. Trout]

Hi all,

It's state very clearly around the VE wiki that the child nodes must be
empty (no VM's, no templates, etc) when they are connected to master.
This makes sense, VMID conflicts are bad.  I'm trying to understand a
little more about the nuances of clustering so that I can have a plan if
things go awry - and to have more things to share when trying to
convince people to use Proxmox VE :).

I'm hoping that the user community might be able to share some of their
own experiences to help current and future Proxmox VE users with
clustering issues beyond initial configuration and setup.

Consider the following use case:

There are two proxmox servers, pm1 and pm2.  Pm1 is runs some important
production services (db, web server) whose availability must not be
interrupted.  However, pm1 needs some hardware fixes.  In order to
maintain service availability, a cluster is created by buying a new
machine, pm2, to be a child node.  pm2 is empty, so it can be used to
create a cluster with no problems.  After the cluster is setup, the
important VMs are live-migrated from pm1 to pm2.  Pm1 is then taken
offline for repairs, and the import servers continue running
uninterrupted.  Management and customers are happy.

What is the best way to manage pm2 while pm1 is offline?  For example,
if a major hardware problem requires the machine to be offline for
several days.  Should one use `pveca -m` to make pm2 the master
temporarily and then again to make pm1 the master once things are back
to normal?

If one has a multi-node cluster (3 or more) does running `pveca -m` on
one node communicate to all other children that it is the new master?
If not, what is an effective means of getting all of the children to
recognize the new master?

Let's say that there is a cluster of 4 machines used by an organization.
 The organization decides to open a second office and wants to host two
additional machines at the second location.  Are there any problems with
extending a cluster in such a way?  It seems like it would be an
effective means for an IT manager at location 1 to manage the entire
organization's servers in one place.  The concern here would be if
something happens to the network connection between the two locations,
would local staff at location 2 be able to directly manage their proxmox
instances temporarily through the child web consoles without causing
problems?

I look forward to hearing how people have handled these or similar
situations.  Please feel free to expand to other scenarios, too.

Many thanks!
Theron

-- 

Theron Trout
Principal, Mellonway LLC

800-614-MWAY (6929)
301-358-5050
www.mellonway.com

Efficiency and Results through Expertise. (TM)