Why should I install OMSA?

[Aaron]

This probably doesn't help your specific request, but I have never 
installed OMSA or any other vendor package on a system that contains 
kernel modules or vendor drivers of any kind. It has always been my 
experience that no matter how much effort they put into writting such 
code, it always manages to destabilize the systems. In my scenario, all 
my machines are commodity servers. If one goes down, I will eventually 
want to know about it, but I don't really care. They are just web/app/db 
servers and the architecture is designed to be fault tolerant. If I had 
a single highly customized non standard build, then I might feel 
inclined to install pieces of OMSA so that I could get better predictive 
data on future failures. With time this won't even matter due to the 
open nature of the hardware monitoring and the continuous improvement of 
the open solutions and standards.

Using this methodology I have only had one pair of servers reboot on 
their own. This was due to a bug between CIFS and Netapp. It was fixed 
and we have been rock solid ever since. The number of servers in an 
organization really does not matter if they are built using a standard 
image, are fault tolerant and their configurations kept in a easy to 
deploy manner. Nagios is more than adequate for monitoring my servers.

That is just my opinion based on my experiences; though I am certain 
others will provide scenarios that this does not work.


Aaron



Gregory Farr wrote:
> Hi all,
>
> I help maintain a 800+ node HPC cluster and I'd like to put OMSA on them
> to help manage, but I need some suggestions to convince the higher-ups
> that its a good idea.
>
> At first there was some resistance about repositories, but this was
> resolved once I told them we could rsync from the main repository to an
> on-site one.  Second is I'm running into other resistance about the size
> of the installation footstep.  It seems that the other sysadmins are a
> bit wary of putting OMSA on the machines due to its package and
> dependency count.
>
> The machines are PE 1950's with DRAC cards and its my assumption that
> OMSA would help us take full advantage of them--especially since smartd
> isn't compatible with 1950's.
>
> Thanks.
>
>