A virtualization blog

Quick post on PSU status

2011-11-06T15:51:00.000-08:00

Hello all.
Today we've had a power outage in our datacenter and i wanted to write something to monitor the fluctuations of the pdus.

Here's the command to check that out

#> environment status chassis

Netapp Monitroing , the quick way

2011-11-01T02:36:00.000-07:00

Hello dear readers , ok so monitoring it's a complex matter i mean you can make it as complex as you'd like but we need to start from somewhere right?

So to me makes sense to monitor 3 basic things:

Is the box up?
How's the CPUs are doing?
What about space ? (specially in aggr0 , you know what happens if aggr0 fills up?)

We will be using snmp , i guess you could do the same thing over SSH but for the sake of this example we will use SNMP v2 (SecurityNotMyProblem).

You can easily enable snmp from either Netapp System Manager or from the console by doing:

(by default netapp comes with snmp enabled (RO) and the comunity string is "public")

Now the next thing is test it up so I'm gonna be using snmpwalk and get from net-snmp tools , you can easily install this in ubuntu by doing..

Great so we have a client that can do SNMP requests and also a server that can respond to these requests , let's test it out:

Great! it's working , it will take you a bit of time to make sense of all of these OIDs but you'll get the idea in further articles.

If you go a download the NETAPP-MIB from http://communities.netapp.com/docs/DOC-1075 , you can open it with any text editor and start to play with it a little bit .. for instance lets do a request for the cpu busy time:

So that interger 5 is the % of time the cpu spent busy ... i know it doesn't look very descriptive at the moment but you could use this information to plot graphs and other things...

We will do that in the next article

to check : http://munin-monitoring.org/

Multistore or Vfilers

2011-10-27T02:55:00.000-07:00

Alrite so vfilers are quite usefull , they're like a logical separator in your filer like a Virtualhost in apache.

To me they're very usefull to people that may be providing some sort of storage hosting etc , or maybe someone that want's to separate environments.

This is what netapp says:

Host multiple customers and/or departments on a single Data ONTAP storage system
Consolidate many file servers, thus reducing cost and increasing staff productivity
Perform simple and fast data migration
Simplify your disaster recovery strategy

This is how you set them up:
Let's create an aggregate first .. raid_dp and 20 drives:

We can check if evertyhing went fine with the following command

Now , very important , add multistore license

License added , so now vfilers live withing a volume therefore we need to create a volume and check its status.

So vfilers either use ip aliases (virtual ips) or ipspaces ... for sake of simplicity i'll use an ip alias.
after your run the create command you gonna go through a similar script like the main setup.. just do as you like there.

And that's it!
Now if you want to run commands on that filer there is a "vfiler run" but to me it's more clear if you switch contexts as shown in the picture below

Once you're withing the vfiler you will see a different promt like "vfilername@storagename" , and you can run virtually any command you'd run on a normal filer with some exceptions.
But for instance .. exportfs , lun , vol are all supported.
I'll try to do something about performance next post .... there's a couple of things that might be handy.

Configuring a netapp HA Pair / Cluster

2011-10-25T02:01:00.000-07:00

So this is very very straight forward , I've done this in the netapp sim you can get from now.netapp.com , or maybe other sources too.

You'll need to download the sim , uncompress it and the run the setup as it is shown in the screenshot below.

For this example we gonna use

Larry(node1) : 192.168.49.222

Moe(node2): 192.168.49.223

I've highlighted the critical bits of the setup , but this is just a 10 minutes setup to you to play .

So i randomly came up with 20 drives but again you can use as many as you like

The next steps are still quite simple , setting up hostname (Larry for this example) , ip address for larry (192.168.49.222) , and one IMPORTANT bit is the ip address that larry will take over when moe goes down (in other words Moe's ip address)

Clustering license is mandatory otherwise the CF features won't be allowed , I've also enabled multistore for latter use:

Here is when you can see my bloogging lack of experience ...hahaha i've selected the text before taking a screenshot ... well it isn't too bad anyways

We'll create an aggregate on larry and this will be our media example that we want to "failover" moe.

So simply aggr create NAME -t RAIDTYPE # (number of disks)

There's also highlighted some commands to check the status of the aggregate.

So what we have until now is , two netapp nodes set up to be an HA pair and one aggregate .

We've added the licenses so now we will need to enable the features , to do so simply do

"cf enable"

If you log in to Moe you'll be able to see that CF will also be enabled if you do "cf status"

We initiate the take over as simple as "cf takeover", you can open another window and ping larry's ip address to see if there's any packet loss.

Once that's done the PROMPT on moe will change saying that it had taken over larry .

You can access the larry by executing "partner" , and then you will be able to access everything.

Well , setting this up is very very simple , have a go at it and comment if you think this was useful.

cf giveback to "giveback" :) , enjoy.

PS:if you need more linceses : http://communities.netapp.com/thread/2463

Next one I'll try to post something about multistore , why to use it etc etc

Why we go virtual?

2011-10-23T16:46:00.000-07:00

I guess one of the main things when your boss tells you we need to go virtual is try to map out where to start , lets say you work in a middle sized company lets say..

Web

2 to 4 dns ervers
5/10 webservers
2/4 databases

Qa/Stg/Dev
10/20 application servers / databases / dns etc etc

Mid office

1 fileserver
1 email solution
1 internal database

Let's assume the common scenario where you are using Apache/Bind/Mysql for your web stack
and Windows 2008 /Exchange / SqlServer for you office things

When i started to use VMware i was very very skeptical , for many different reasons such as : it isn't going to perform , i don't trust vmotion , i don't trust the back-end storage , i don't trust the memory distribution and reservation , i don't trust other apps sucking all the memory up etc etc
All these concerns are very valid you need to be sure where you are getting into , at the end of day whatever hypervisor you go with it will be the backbone of your infrastructure.

To me it seemed very logical to start with something that wouldn't affect either production applications nor business as usual systems , so i decided to virtualize our QA environments.

It's kind of simple to see that it's always easier to deal with internal developers if something goes wrong than with either customers or excel users that want to access a db etc etc.

Next post:
p2v , kernel settings for unix , vmware tools , and a few words on choosing the back-end storage protocol and appliance.