I had sent a previous mail, but it was too large for the list.
I am doing similar to Mark - I have a few perl scripts pulling stats via snmp and cli perf commands. These get pushed into graphite, and I am using graphitus as a dashboard to view overall stats.
Graphite is fairly powerful - it allows queries like "top 5 average write response times" over all disks to be easily identified.
Chris
On 2013/08/01 10:00 PM, Weber, Mark A wrote:
Our environment isn't as large as yours, but I wrote a few powershell scripts that pull metrics and push them into graphite and mssql. Things needed for billing reconciliation go into sql and performance/trending into graphite. Using graphite's render API it is fairly straight forward to roll-up like things into a single graph - like sum all IO on each volume on a specific aggregate and divide by the sum of the used space to get the IO density.
mark
-----Original Message----- From: toasters-bounces@teaparty.net [mailto:toasters-bounces@teaparty.net] On Behalf Of Ray Van Dolson Sent: Tuesday, July 30, 2013 1:11 AM To: toasters@teaparty.net Subject: Metrics!
We have a growing deployment with ~25 filers and about 750TB of usable space. Most of it is VMware datastore space exposed via NFS, but some is general purpose NAS file sharing.
We've been using OnCommand Core (formerly DFM) to handle reporting and such, and for the most part it's worked pretty well.
I need to start pulling together some metrics for the environment as a whole, both to benefit the Storage Team who manages the devices, but also to be able to pass along to my bosses so they can get a basic grasp of how things are going.
I'm curious how those of you with deployments similar to mine or larger hvae dealt with this? What sort of things are you reporting on, and how do you handle the "roll-ups" to summarize many (in our case 200+) volumes into something meaningful? Are you generating reports natively from OnCommand Core, or pulling the data out into another tool? Do you find the OnCommand DataSets feature useful? We haven't really leveraged it properly I think.
My thought is to present a sort of physical view of our environment showing roll-ups of storage capacity and performance (probably CPU, IOPS and latency). How to give a snapshot overview of the whole environment is what I'm not sure on.
I'm thinking using the DataSets will help me organize our various types of data and then I can report on that.
I'm not sure, however, if I should spend my time trying to get the right reports out of OnCommand Core directly or if I should just start extracting the data to another tool (which?) or into a database where I can home-brew up what I'm after.
Thoughts/experiences?
Thanks, Ray _______________________________________________ Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters