Thursday, January 30, 2014

Nimble Scale-Out - The Curtain is Drawn - Part II - What's a Group?

Hi Friends,

Today I'm going to go into a bit more depth when adding 2.x arrays to an existing 2.x group.  Remember I showed you how easy it was and I said, "Well, it's a little more involved then that."?  Today I thought I'd show you what's involved and maybe some other bits.

So if you're upgrading a 1.x array to a 2.x array you need to start thinking about your network.  Since this is going to be a cluster some of the individuality that you had with 1.x will be lost when you upgrade to 2.x and join a group.  Now I just used the word "group" a couple of times, and you're probably thinking, "Okay, what's a group?"  Well, it's just that, it's a bunch of arrays(up to 4) joined together to form a group.  And once an array joins a group, it gets all the rights and privileges an array gets at this point.  It's kind of like moving into a fancy gated community.

Sounds pretty exciting eh?  So how does an array get ready to join this exclusive club?  Just like in real life, networking my friends!

Let's take a look at this networking GUI.  I've already added "whenyou" to the same group "wenster" belonged to.  If you haven't guessed it yet, these were named by my buddy Wendel. :-)  Now take a look "iSCSI Host Connection Method".  This is new concept if you've only worked with 1.x.  If you upgrade an array from 1.x to 2.x most likely it will have "Manual" and "Enable rebalancing" not selected.  If you want that array to remain a separate entity, that's okay, just leave it alone.  It can happily live out it's life being a loner and serving up data as a 2.x array without ever being part of a group.  But if you want to live in the fancy community with the dog park and the swimming pool you're going to need the networking and learn to work with others.

Also notice the "Virtual Target IP", another new concept.  Now that you're networking with others, you need an easy way for data to get from point A to B.  I still have data IP's for "wenster" and "whenyou", but now the Virtual Target IP helps sort what data needs to go where automatically!  I know what you're thinking, "How do you keep data from going over the wrong pipes?"  That's an EXCELLENT question!  And all I'm going to say is Nimble's got you covered!  I'll probably write about that in a day or two.  I will say when you populate a group, you're going to want to switch from "Manual" to "Automatic" and select the "Enable rebalancing".

Now, back to networking.  Remember I said that before you start adding arrays to group, think about networking and here's why.

When you try to add an array to a group the array going to try and keep you from making any mistakes.  Pay particular attention to bullets 1, 2, 3, oh heck just read them all!  We wouldn't put them there if they weren't important.  Remember, don't be like Neil and not read the pop ups, they're not there to just annoy you.

Done reading the bullets?  So what we've got is two arrays, "wenster" and "whenyou" and both now belong to a group called "hulk."  Hulk will now take over much of the network identity of the new array that's been added.  Is there a downside to this?  No, not really, BUT you need to remember to prepare for the changes.  It's all about the networking my friends!

So after you've clicked Finish with out reading any of the bullets you will hopefully get this message box.  What!  Wait a second, go back and read those bullet points!

As stated in the earlier bullet points, you'll reconfigure your applications to use the "hulk" or group target IP address and the management of the individual arrays is handled by the group management IP address.

You've just taken the first step to a bright and brand new world.  It's like in the Wizard of Oz where it goes from black and white to color.  Black and white is cool, nothing wrong with it, but WOW what a difference the world looks like in color!

Until Next Time

Wednesday, January 22, 2014

Never Fear - Brain & Wendel Are Here!

Hi Friends,

Starting to go through Brain & Wendel withdraws?  How about a dose of Brain & Wendel Live!  Yes that's right here's our second live webinar where we discuss disaster recovery in a virtualized world.


Thursday, January 16, 2014

Brain & Wendel Episode 12 - Oracle Best Practices with Iron Tom!

Hi Friends,

For all you Oracle fans out there, this week Brain and Wendel talk with Iron Tom.  Tom is a member of our team and brilliant Oracle and Linux guru.  If you've got questions about Oracle databases tune into this weeks Brain & Wendel Show and enjoy!  Also check out Tom's blog for tons of great Oracle content:



Citrix XenDesktop 7.1 on Microsoft Hyper-V 2012 with Nimble Storage and Cisco UCS

Hi Friends,

We've all been hearing a lot more buzz about Microsoft Hyper-V 2012.  Will it be the next Hypervisor of choice?  Only time will tell, but if you're like me and want to hear about and read all the cool new VDI architectures you'll enjoy this one!  Citrix Consulting tested Hyper-V 2012 with the new version of XenDesktop 7.1 along with Cisco UCS and Nimble Storage.

The cool thing is that if you're looking to implement something similar, this document will be a huge help!  This is a complex machine they've built and have done a great job at it.  If you know me, you know I love mechanical watches.  Just the accuracy and art work that goes into creating a mechanical time piece is like an evening at the symphony.  This architecture might not make music, but watching all of these complex parts come together to build a functioning system is quite impressive!  As I continue to dissect this whitepaper I'll be blogging on it more, but I wanted to share it.

Here's where you go to get it:

Until Next Time!

Tuesday, January 14, 2014

How to Setup MultiPath (MPIO) with Nimble Storage and Citrix XenServer

Hi Friends,

Here's a follow up to my Citrix XenServer blog that showed you how to attach Nimble Storage to XenServer.  This article will show you how to set up MPIO with Nimble and XenServer.  It was written by a colleague of mine, so a HUGE thanks to Klaus Kupfer!!

 1.  Configure each Data Interface on the Nimble Array to a different subnet.

2. Configure Multipathing on the XEN Host
      a.       Put XEN Server in Maintenance Mode

 b.      Access XEN Server Properties


c.       Enable Multipathing
d.      Exit Maintenance Mode

3. Check that the XEN Server NIC’s are in the same subnets as the Nimble Data Interfaces

4.  Edit the /etc/multipath.conf File using console
device {
vendor                                    "Nimble"
product                                  "Server"
path_selector                         "round-robin 0"
path_grouping_policy           group_by_serial
rr_weight                                priorities
rr_min_io                                10
path_checker                         tur
failback                                   immediate
no_path_retry                        20

5.  Reboot the XEN Server

6. Create New SR 
     a.       Obtain initiator IQN
     b.      Configure Volume on Nimble on Nimble array with all desired settings and LUN initiator groups

c.       Choose New SR...


d.  Choose "Software iSCSI“ as Storage Type

e.       Configure a name for the SR

f.        Enter the iSCSI Discovery AddressThen click first  "Discover IQNs“ and then "Discover LUNs“ The configured Volume should appear in the list

 g.  Finished

Until Next Time!

Thursday, January 9, 2014

Nimble Scale-Out - The Curtain is Drawn - Part I

Hi Friends,

In today's blog I wanted to dive a bit deeper into the new features available with Nimble Scale-Out.  When you get your first storage array it's great, you've got your GUI or CLI and you're ready to go!  But then you get your second and it too has a GUI, which is okay.  Then a 3rd with another GUI, which is starting to get confusing.  Let's top is off with a fourth which, it too has it's own GUI.  Did I just delete the volume on Array 1 or Array 4?  Ugh oh....

You get my point, when you have multiple GUI's open, it starts to get a bit confusing.  The cool thing about Scale-Out is you have one GUI, and you don't have to download a separate tool to administer your cluster!  Yep, it's all built in.  Basically you elect one array to be your leader and it's the first one to join the cluster.  As others join they give up their ability to have a separate front end GUI, since you'll be using the "leaders" IP address to administer from 1 to 4 arrays.

What if you want the ability to have a single pane of administration glass, but prefer each array to be their own entity.  You're not ready to start striping volumes or use some of the newer features.  That's okay!  Scale-Out isn't an all or nothing upgrade, you can jump into the deep end or slowly get your feet wet at the shallow end of the pool.

How about some pictures?

Here's we've got a single array called Anooba.  It's group is called t1 (I'll get into groups a little later on) and it's the only array in the group.  We'll click on the Add Array to Group button and....

We get a box that shows us the available 2.x arrays available on the network.  Pretty cool huh?  Click on the one you want to add, here I select Bantha.  Click on the Add button.

What?!  A password dialog!?!  Yep, believe me, this is a GOOD thing, you don't want arrays magically appearing in your cluster.

After you enter your password for the array you're trying to add to the cluster, click Connect.

Check it out!  Both arrays are there.  Was it really that easy?  Well, I cheated a little.  There are some networking considerations that you need to be aware of when adding arrays to your cluster, but that's for another blog.  Hey, I gotta get you coming back!

Until Next Time,

Tuesday, January 7, 2014

Now Introducing Nimble Scale-Out! This IS The Storage You're Looking For!

Hi Friends,

Today I'd like to introduce a super cool new product put out by Nimble Storage.  Scale-Out, Nimble's  next gen version of storage.  Yeah yeah, I'm sure you've heard that a lot and you're thinking, "Blah blah blah, show me the money!"  Ever watch Looney Tunes?  You're probably thinking Scale-Out is like this:

Well my friends in this case you'd be wrong!  Scale-Out is much more like this:

And the great thing about about it?  You can download it today!  It's a release candidate that's available for download through InfoSight.  And in most cases will work on your existing hardware.  AH HA!  You're thinking, "Most cases, meaning not my case."  No, it'll probably work on your hardware too, but there's always a small percentage so it's best to check with your account team.

Okay, okay so Michigan J. Frog aside, let's talk about what's so fab about it.

A large differentiator between Nimble and other storage is it doesn't rely on spindles for performance.  CPUs are cheap, we use them for our performance.  Disks are just a way to store your stuff cheaply!  You need more storage, add a shelf of disks.  Now what if you want the ability to administer your storage from a single GUI?  Stripe a volume across all of your storage controllers for smoking fast performance?  The ability to evacuate an entire controller for maintenance or storage replacement with NO downtime?

Yes, friends, you get this, the car polish AND the set of knives....  Sorry couldn't help myself.

Want to know what else is cool?  The GUI hardly changes too!  When I did my first upgrade I thought I had done something wrong because it looked too much like 1.x Nimble.

Here's a 1.x system:

Here's a 2.x system:

At first glance, not much difference huh?  I kept waiting for the other shoe to drop, but it didn't!  Now of course there are differences in there and some caveats before upgrading which I'll go through in another blog, but on the painful scale I don't think it was even a stubbed toe!

That's all folks!

Monday, January 6, 2014

Neil Versus PVS - Part IV - Multiple Sites with KMS

Hi Friends,

Here's an interesting one that I ran into while working with multiple PVS Sites.  After the fact it seems kind of straight forward, but at the time I really had no idea what was going on.

If you take a look at my architecture I have three sites and two PVS servers gong to each site.  Ignore "Site" I was using that for testing.  :-)

The thought was with multiple PVS servers, you'll probably start having multiple sites too.  It's also easier with house keeping when you start collecting a lot of servers, volumes, vdisks, etc.  My vDisks are all sitting on a CIFS share so I don't have to have multiple vDisk images, which is VERY handy.  There's only one small problem and I probably wouldn't have discovered it for awhile, but thanks to KMS, my life got very interesting!

I decided to log into UCSPVS3 which was my third PVS server.  I loaded up the Provisioning Services Console and everything looked normal.  I could even manipulate vDisk images mapped to Site 1!  Now that's cool!  Now here's where the problem began.  I went to change the vDisk from Private image mode to Standard image mode and I got this error message:

Unable to map the vDisk.  Mapping was denied by the Server.  Huh?  And when I turned KMS off as an option under vDisk Properties, it switched from Private to Standard image just fine!  So what the heck?

You might be saying, "Just use the Site your vDisk is assigned to Neil!!"  Well, yeah, but isn't the point of having it on a CIFS share is I can share it from anywhere?

Well, not exactly.  So here's something I found out.  When you have the KMS option enabled in the vDisk Properties, the vDisk momentarily gets mounted up as a volume in Microsoft Disk Management on the PVS server you're trying to switch from Private to Standard image.

Here in lies the problem....  I was trying to mount up the volume on a PVS server that didn't have the vDisk assigned to it!  As I said earlier, the only reason I had this problem is with KMS enabled the volume gets momentarily mounted.  So if not for KMS, I probably wouldn't have bumped into this unless I tried to mount up the vDisk.

After the fact it seemed so simple and straight forward, but sometimes when you're lost in the forest, it's hard to see the trees!  :-)

Until Next Time!