Running EMC ScaleIO In A Nested vSphere Environment

Hi,

the first post includes the basic setup for 3 nested ESXi Hosts, in my laptop I only have 16GB of ram so that won’t work, so this post will describe the installation on only two ESX hosts, how can we do it if we need a minimum of 3 ScaleIO entities, well? im doing something here that isn’t supported in real life, basically, one ESX hosts will have one ScaelIO appliance and the 2nd ESX hosts will have 2 ScaleIO appliances (that’s the part that isnt supported), this will reduce the overall memory footprint while still “work”

this post will also include the setup of the ScaleIO packages + installation without the usage of the wizard that was included in the first post!

the screenshot below shows the configuration on the 2nd host, basically, it will run 2 ScaleIO ova appliances, each on the 10GB file, and each one of the 50GB file will be used as the “raw” “drive” that will be attach to each OVA.

in this case, I used one IP for the OVA that reside on ESX1

and this is how it looks when im using the ESX 2 iSCSI initiators to point to the iSCSI ScaleIO appliance

and two appliances on ESX2(one per appliance for the two that will run on ESX2)

and of course, this ESX hosts will now need to point to 2 ScaleIO appliances as well

but as we mention, this will not happen in real life, in real life, each ESX host will point to one ScaleIO appliance (the one that it runs)

like in the other post, we will need to copy the ScaleIO installation file to each appliance (opt/scaleio/siinstall) se the permissions properly (777)

now, we will need to extract the installation files (in each appliance)

ok, this is where stuff is starting to get different than in the automated install, as oppose to running the install.py script, we will install the RPM packages that are located in the “packages” folder

note that in the “3” node configuration, we will install the MDM cluster on two nodes and the Tie Breaker on the 3rd node, basically, the primary MDM goes to ESX01, Secondary node will goe to ESX2 OVA1 and the Tier Breaker will go to ESX02 OVA02

just as a reminder, it’s not a must to install the SDC and the SDS on the same ESX server if don’t want to but the majority of the real world use cases will do just this because it make more sense for your ESX host to provide compute (SDC) and storage (SDS)

ESX01 (MDM, SDC, SDS and the SCSI target package)

ESX02 – OVA1

ESX02 – OVA2 (Tie Breaker – NO MDM!)

Installing ScaleIO Components

1. Install the MDMs on nodes IP1 and IP2, using the command:
rpm –i <MDM package>
2. Install the Tie-Breaker on node IP3, using the command:
rpm –i <TB package>

Install the SDS on all nodes (IP1, IP2, IP3), using the command:
rpm –i <SDS package>
4. Install the SDC on node IP1, using the command:
MDM_IP=IP_Cluster rpm –i <SDC package>

ok, we now need to prepare the MDM’s on the different appliances

ESX01

ESX02

ESX03

Configuring the MDM’s / Tie Breaker

ESX01

ESX02 (performed from ESX01 !!)

ESX03 (performed from ESX01) – installing the Tier Breaker

Switch to cluster mode (performed from ESX01)

entering license key

define protection domain

adding capacity (the local 50GB drives per appliance)

add the newly created 50GB drive to the appliance as an additional drive..(need to repeat this on ALL the ESX hosts)

and then format it as a VMFS datastore (eager zeroed thick)

the next stage is to add these 50GB drives to the appliances, each 50GB drive goes to another appliance

you need to reboot the appliances in order to see the newly added drives, after the reboot run fdisk –l to see if the 50GB drives have been added (run the command on all the appliances)