On 11-09-22 10:41 AM, Nick Khamis wrote:
> Hello Everyone,
>
> We have almost setup a working prototype of what will be our
> production cluster. A few simple question I have are:
>
> i) We begin the installation by creating hacluster:haclient. How bad
> is it to proceed with the installation as user haclient
> of course sudo'ing the make install process? Do we have to depend so
> much on root for installing and running the stack?
>
> ii) Does the complete stack (RA, glue, corosync, pacemaker) have to be
> installed on every machine in the cluster
> (load balancers, real servers). What we have is:
>
> lb1.whaterve.com&  lb2.whatever.com
> These contain ldirectord, and therefore would need RA, glue, corosync,
> and pacemaker? Is that correct?
> Since the entire stack is installed on these machines, we can also use
> the to load the RA primitives for
> mydrbd1&  mydrbd2?
>
> mydrbd1.whatever.com&  mydrbd2.whatever.com
> These are real servers that run mysql on "network Raid 1" using drbd.
> Do we need to install the entire cluster stack
> on those nodes as well? If not, what is the bare minimum that needs to
> installed on these machines, in order to be
> included as part of the HA cluster?
>
> Thanks in Advance,
>
> Nick.
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>
I've installed several clusters and while I'm not an expert, I would say 
you're pretty much going to have to install the entire cluster stack on 
any node that is going to be a member of a cluster. The amount of space 
they all take up is minimal so I wouldn't see the benefit of trying to 
not install one piece. Install the entire stack as the instructions 
indicate and you'll have all the pieces you need to make a cluster 
operative.

I've not tried installing the software as a user other than root. The 
default install path is going to be in /etc/ha.d or /usr/etc/ha.d so 
having write access to subdirs under /etc and /usr is  necessary.  Also, 
don't forget that your RAs need to do things like bring up a virtual IP 
- if you install as haclient this is not likely going to work. Install 
as root. If it was better/easier to do it as another user then the 
instructions would have you do that.

You do not need to install cluster stack on "real" or backend servers 
just the nodes that are actually part of the cluster.
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to