Slashdot Mirror
GUI-Based Asset-Tracking Tools For a Datacenter?
toruonu writes "How do you keep track of what's in your datacenter, where it is, what it's connected to and what is it doing right now? I mean I have built a datacenter from scratch over the years and I have machines from Sun, IBM, HP, Supermicro. I have machines that are simple workernodes and machines that are heavy grade storage consolidation machines. Then there are tens of switches, some for interconnect, some for management and don't get me started on the UPSs etc. So how does one keep any kind of decent track of such a system as the current form of twiki pages with various tables just doesn't cut it anymore and I'm looking for a freeware solution that could actually show me a visual representation of the various nodes in the racks, their connections and dependencies. Just to give a simple example, if I'm going to disconnect UPS #3 right now and swap switch #5, which machines should I even consider taking offline?" (The best-looking such system I've seen was being used at OSCON at a display booth for the Open Source Lab, and I think it was home-grown. Anyone who can shed light on that system?)
Rows for hosts, columns for PDU, switch and console ports. Additional rows for asset tag information, unit manufacturer, model number, serial number. Last row for notes on the system, e.g. any historical hardware issues that may be relevant.
Might be close enough, I guess...since that's essentially what you want to do, map some aspects of the "electronic mind" that's under your care.
Wikipedia seems to hava a list perfectly adequate as a starting point
http://en.wikipedia.org/wiki/List_of_mind_mapping_software
One that hath name thou can not otter
Not sure if it meets his needs exactly, but I've used http://racktables.org/ in the past and it's worked well for keeping track of a small-ish datacenter (about 400 sqft with 7 full size racks and a couple dozen servers).
"People who think they know everything are very annoying to those of us who do."-Mark Twain
A couple of weeks ago there was this company called Bright Computing, that was pitching their software called Cluster Manager, which looked very cool as they had an excellent interface and could keep track of everything from machines, to routers, switches, power usage etc. Something to look into.
http://www.brightcomputing.com/
I use Nagios for that kind of thing. Don't get me wrong, it isn't "perfect" at it, but it does a decent job once setup. If you use parenting in the configuration files, you can click on "network map", and immediately see each hosts' dependencies. And IIRC there are comment fields that you can write misc information (such as rack position, switch position, model, make, etc)... And it's free...
If a man isn't willing to take some risk for his opinions, either his opinions are no good or he's no good
This problem has several paid solutions, all of which work fairly well, and make maintaining a data center the job of one person, instead of 20 people looking at a spreadsheet and log files. I haven't found an open-source package that is nearly as competent as the integrated solutions offered by HP, IBM and others. Warning: sticker-shock is included. Bonus: PHBs like looking at pretty pictures, and all the commercial tracking software produces pretty pictures. Your PHB looks like a super-hero to his PHBs, and you become an invaluable asset to your PHB for making him look good.
Those who can, do. Those who can't, sue.
I don't know your solution, but I can tell you it will involve automatic network mapping and polling of services. You need to find a solution that relies on human input as little as possible. Otherwise documentation gets out of date, no longer trustworthy, leading to lack of incentive to update it, ...
With a big budget, I'd go for RFID on everything, with local readers doing triangulation. That's the only way to really track physical objects. Add that to the maps that network discovery makes and you've got what you need.
On the one hand you take life too seriously, and on the other, you do not take playful existence seriously enough. Seth
The system used to visualize the OSUOSL lab is called Rackview.
http://rackview.sourceforge.net/
http://flux.org.uk/projects/rackmonkey/
This is what I use to keep track of the racks in my work's facility. It allows you to put in a whole lot more than just simple rack location. It's a wonderful tool.
Seriously, we looked all over for something, but nothing fit. Grant it, we did this back in 2000-2001 timeframe. We setup a mysql database, and wrote up a website with php which was the interface. We scanned in floor plans of the buildings and setup an image clickmap for all the cubicles/locations on the floor plans and had them all point to a unique location_id. The location_id's were one of the keys in the datbase to track the hardware.
So you could litterally navigate to a particular building/floor, and then click on the cube/location and it will then show a list of all the equipment in that area. You can add new hardware to that location or click on a piece of hardware and view its information (CPU type/speed, hostname, IP address, MAC address, RAM, etc..) and if it moved to a new location, you click on a "move" button, and it opens up the list of buildings/floor plans and you simply navigate and click on the place where it moved to, and then update the database record.
We add some more sophisticated features like barcodes to cubicles and to the systems themselves and you can go around with a barcode scanner hooked up to a laptop and simply scan the barcode on a cubicle, and then scan all the barcodes of equipment in the cubicle and it would automatically associate all that equipment with that particular location (and if it was a new piece of equipment, it would open the form to add the equipment into the database).
We were all warned a long time ago that MS products sucked, remember the Magic 8 Ball said, "Outlook not so good"
It does automated network scanning for assets, asset tracking, alerts and notifications, SNMP data collection, reporting, and yes, it has a .svg based network mapper that you can customize with your own graphics. It comes with MIBS for hundreds of devices, but you can easily import your own MIBS for unsupported devices. It's open source, of course. Nagios is just a bunch of disassembled parts. You have to wire it together for each device. Adding new devices is a pain: you have to install the Nagios monitors on each new device. Nagios does not speak SNMP! OpenNMS does speak SNMP, and it will autoscan networks for devices, and devices for capabilities. Adding thousands of devices at a time is a snap. Plus, OpenNMS uses a modular architecture that scales well. We use it on a network consisting of over 2,000 clients at 50 offices, 30 IBM Blade servers hosting hundreds of VMWare virtual hosts, and innumerable network devices, printers, etc.
- None can love freedom heartily, but good men; the rest love not freedom, but license. -- John Milton
The project you saw at OSCON was called RAIV (Rack And Inventory Viewer). Unfortunately it ran into a dead-end and is currently vapor-ware. Currently we're using an internal CakePHP webapp for basic inventory and customer tracking, but its very buggy and lacks many features.
We are in the midst of working on a completely new project that will cover many of the problems mentioned in this article and beyond. Think of it like an open source datacenter management webapp and backend. Its still in the planning stages, but the intent is to have a plugin based system where you can use the inventory plugin, DNS/DHCP plugin (to replace maintain), virtualization management (deployment and console access), etc all in one interface. The idea is to create an admin interface and a customer interface so that they can access and see information about their environment. We're far from having a demoed project but we hope to have something soon.
I use Zabbix to monitor everything. It will monitor just about everything out there and isn't just limited to SNMP like OpenNMS. It is much easier to install, configure, and maintain then Nagios and it has much prettier graphs and management tools then Cacti.
The reason why Zabbix stands out from the others to me is because of how well it functions in the server room for monitoring, alerting, and self healing plus when management walks by they are always impressed. The display that sits on the wall is visually appealing to them when they see the graphs and colors and since that system is set up for read access only they can drag time lines around, see other graphs, print reports on trouble systems, and they can do this on their own (aka: they don't pester me for the information!). Meanwhile, Zabbix is off and monitoring things like the DHCP server so that when a MAC it doesn't know shows up, I get a complete nmap scan of the system (tools are supported out of the box, but this is something you setup manually in the discovery section), and the systems activities are tracked and monitored until the box is configured as a trusted server. Zabbix watches things like a proprietary (ugh) program that is known to crash a few times a week and when it does crash, Zabbix flushes the logs, and restarts the program for me; I just get an email "The program crashed; I fixed it." I even have Zabbix monitor SMART information on the hard drives so I can track everything down to the hard drive serial numbers, temperature, and prefail states. Several of the UPS's are fairly intelligent and work well with the OS, so I have Zabbix monitor those (those few that don't even have a port to connect to the computer I still track manually). It does everything from problem finder, to healer, to network watchdog.
The one thing it doesn't do well is the automatic population of server data even though it has manual entry fields for Server Serial Number, MAC address's, ect, ect, ect. However, it was trivial setting up Zabbix to run a script that gathers that information up and dumps it into the SQL fields for that system.