I'm working with an environment that spans multiple data centers/sites and consists primarily of HP ProLiant servers (G5-G7) running Linux. The mix is 30% RHEL/CentOS, the rest are Gentoo :(. I also have a few dozen virtual machines running back-office and Windows servers on VMWare ESX hosts. I run OpenNMS to pull SNMP data from the various server nodes and networking devices. While OpenNMS works wonderfully for up/down, thresholds and notifications, it's native handling of traps is a little rough and the graphs are not particularly pretty. I use Orca/RRD graphs for performance trending and nice graphs.
I'm tasked with inventorying the environment and wanted to come up with a clean way to organize server information. Since my environment is mostly HP, I've been playing with HP Systems Insight Manager as a way to extract server data and to deploy HP health/monitoring packages and firmware. The Gentoo systems eventually have to be converted to CentOS, so getting a quick assessment of what hardware is where would be great.
Although I've read through a few hundred pages of HP manuals, I'm having a difficult time understanding how to get HP SIM to do what I want, though. My main problems are:
I have about 40 subnets to deal with; 98% connected with private lines to facilities across the globe. I don't want to initiate an HP SIM discovery only to pull back every piece of intermediate networking hardware and equipment from all of the locations. I'd like this to focus on the servers.
I have OpenNMS configured to accept traps. I don't want HP SIM to duplicate that effort. It seems like the built-in software deployment tool wants to overwrite the trapsink parameters for the systems it encounters during discovery.
I have about 10 administrative username/password combinations in use across this infrastructure. Is there a more efficient way to get HP SIM to do the discovery or break discovery into manageable chunks?
In terms of general workflow, do people typically install the HP Management Agents during the initial OS deployment (e.g. kickstart post script) or afterwards from HP SIM?
Is HP SIM too thick/fat to be an inventory tool? I can't tell if it's meant to be used standalone or alongside other monitoring products.
Since the majority of the systems I'm trying to track are those running Gentoo (in order to plan the move to CentOS), is there any way for HP SIM to extract system model information from them ( like dmidecode)?
I have systems here where I may have an SSH key established, but not direct user or login access. Is there a way for me to import an SSH private/public key pair into HP SIM to reach out to the servers that can't accept standard credentials?
There are a handful of sites where I have inconsistent access or have a double-NAT situation. I may be able to poke a server, but it may not be able to find its way back to the management system. Is there a workaround for this?
The certificate configuration for HP SIM seems complicated. What is the preferred setup for trust between systems?
I'd also appreciate any notes or recommendations to using this product. Or if there's a better way to do this, I'd like to know.
