Simply import either of the host template XML files into your Cacti installation. Place the included xml files in cacti/resource/snmp_queries/
Use the “HP LeftHand Cluster” template only for Storage Clusters, and enter in the VIP (Virtual IP) associated with the cluster for the hostname.
Use the “HP LeftHand System” template for all individual storage nodes. I would recommend adding in things like Interface Statistics, RFC 1213 stats, ucd/net Load,CPU, Memory etc. I decided not to add these into the host template I'm providing because I didn't want all the dependencies to screw up other people's Cacti Installations.
Technically, the Data Queries being used in the Cluster host template can be used against ANY storage node, but due to the nature of how NSM-CLUSTERING works, all of this information is available anywhere you look, so you could end up graphing Volume usage statistics for the same volumes in multiple places. It's best to just point to a single VIP, and let the clustering logic take care of getting Cacti (and all your other storage needs!) to where they need to be.
Post your comments and troubleshooting questions here: Host Templates for HP LeftHand Storage System
These templates have been under development for quite some time. They are still not done, but I'm getting them out here for the community to use. There are MANY problems with the “HP - LeftHand - RAID” data query. Pretty much every metric is wrong. This is not my fault, but HP's. Their SNMP values for these metrics are either completely incorrect or increasing at unprecedented rates. I have a support case open with HP to determine what (if any) CDEF needs to be applied to make sense of the values being reported. The I/O Latencies being reported in both the “HP - LeftHand - Volumes” and “HP - LeftHand - Cluster” are also suspect. What is reported in the CMC for latency versus what is reported via SNMP is very different. I played with a couple different CDEF's to 'guess' what HP might be thinking here but gave up and left it alone. The MIB file only explains that the latency is being reported in 'ms', and it's a COUNTER type. I've set a limit on the latencies to 3000 ms, and in my environment I always see breaks in the graphs, but this is better than seeing a 10P spike that blows all the other data out of the picture. The throughput graphs are accurate. The IOPs seem to be accurate. The “Creation Time” field on the “HP LeftHand - Volume Usage” graph template always reports HEX … for me. This is because the MIB file isn't being used during the query by Cacti. I don't know how to fix that.
While working on these templates, I noticed that per-initiator stats are now available through SNMP. I decided not to dive into that for this version, as our environment as a LOT of initiators and it would be massive overkill to use. For smaller organizations with only a couple initiators talking to the LeftHand clusters this is probably something that would be desirable. Let me know if you're interested, and I'll find some time to implement it.
If you've been around the block with Cacti templates, you'll most likely see some similarities between these templates and others that are out there. I can't even begin to imagine where I've stolen all the bits and pieces from. Off the top of my head … Thanks to the netapp template team/person, BSOD2600 and his RFC1213 template which I've used as the standard for creating all my Data Queries, Mark Round and the Solaris/Linux iostat templates … and the folks at CU for putting up with me :) If something I've done is blatantly stolen without credit, please let me know and I'd be happy to give credit where it is due!