Site icon

A closer look at ECS (Elastic Cloud Storage) running on vSAN

This week I had the opportunity to take a closer look at ECS, the Elastic Cloud Storage product from DELL-EMC. ECS normally ships as a physical appliance, composed of multiple nodes/servers, but there is also a community edition available which is FREE with no time limit for non-production use. You can download it from here, and deploy it across VMs. Now I’ve already been looking at various ways to provide an S3 object store on top of vSAN, such as MinIO and Scality. What is interesting about ECS is that not only can it provide you with an S3 object storage, but it can also provide HDFS for Hadoop, Swift and NFS for file services. I decided to deploy this on my 4-node vSAN cluster, and I rolled out 4 x ECS VMs, one per vSAN node. To each ECS node, I provided a ~100GB VMDK, which was instantiated as a RAID-5 volume in my vSAN cluster. While the complete set of steps for doing a deployment are located here, I condensed down the steps involved in rolling this out on vSAN.

Minimum virtual hardware requirements

Once I had deployed my 4 VMs, I used DRS affinity rules to pin each VM to its own ESXi host in the vSAN cluster. All of my disks were RAID-5, but there is no reason why you should have to use this configuration for the disks. You can play around with stripe widths,as well as mirroring instead of erasure coding to see what works best for you.

Ensure you set the memory to the correct amount. Step1 of the deployment will fail with an error if it finds less than 16GB.

User requirements

Network requirements

If you do not disable rpcbind, step1 of the deployment will fail with port conflicts.

Misc. requirements

Install the software

Now use git to download the software as follows:

[chogan@ecs-node-1 ~]$ git clone into 'ECS-CommunityEdition'...
remote: Counting objects: 4394, done.
remote: Compressing objects: 100% (14/14), done.
remote: Total 4394 (delta 2), reused 3 (delta 0), pack-reused 4380
Receiving objects: 100% (4394/4394), 4.62 MiB | 1.57 MiB/s, done.
Resolving deltas: 100% (2771/2771), done.
[chogan@ecs-node-1 ~]$ cd ECS-CommunityEdition/
[chogan@ecs-node-1 ECS-CommunityEdition]$ ls
bootstrap_plugins   contrib   Jenkinsfile     tests        docs      license.txt  README.rst    tools  examples  patches      release.conf  ui
[chogan@ecs-node-1 ECS-CommunityEdition]$

 Modify the deployment yml file

A sample deploy.yml file is shipped with the software. You need to copy it to the ECS-CommunityEdition folder and modify it accordingly.

[chogan@ecs-node-1 ECS-CommunityEdition]$ cp docs/design/reference.deploy.yml deploy.yml
[chogan@ecs-node-1 ECS-CommunityEdition]$ vi deploy.yml
Various fields need to be modified to match your environment (these are typically labeled required in the yml file). Here is a comparison of the yml file from before and after the edits for my setup (the /dev/sdb reference is the second disk (104GB) device on each VM):
 license_accepted: true


  ssh_username: chogan
  ssh_password: xxx
  ansible_become_pass: xxx


 - /dev/sdb

 - name: sp1
 - /dev/sdb

 Start the deployment

To start the deployment, run the script using the deploy.yml file that you just edited to reflect your environment. You can see that this is not being run as root, but with an admin user called chogan who has sudo privileges – note that the superuser password is prompted for on a number of times during the bootstrap.
[chogan@ecs-node-1 ECS-CommunityEdition]$./ -c deploy.yml -y
>  ECS Community Edition Install Node Bootstrap 2.5.2r
>  ECS Software Image emccorp/ecs-software-3.1.0:hf1
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Run step1

Once the install host has rebooted, log back in as the admin user, and run step1. I’ve truncated the output here as there is simply too much of it to display.
[chogan@ecs-node-1 ECS-CommunityEdition]$ step1
[sudo] password for chogan:
ecs-install> Initializing data container, one moment ... OK
ecs-install> Applying deploy.yml
This step1 does a number of things such as setting up ssh keys, sudo password verification, checking firewalld is running and adding all nodes to the trusted zone. One thing I wasn’t expecting was that the script also sets new hostnames on the hosts. It seems that it is using moon names, such as luna, phobos, deimos and io. This was very confusing to begin with, but rest assured it is expected. It also updates DNS records as it needs to resolve both the short-name and FQDN of the host. The script also verifies that the minimum requirements from a virtual hardware perspective are met. It also verifies that the block device (/dev/sdb) that we placed in the YAML file exists. These are formatted and mounted during this step. This is also the point where it checks whether any ports needed by ECS are in use, which is why we disabled rpcbind earlier. When this finally completes, the ECS services are started.

Run step2

Once step1 is completed, it is now time to start step2. This step takes care of licensing, creating storage pools, VDCs, replication groups, namespaces, as well as initial users. Here is an example output from a step2 run taken from my deployment. Again, I’ve truncated the output as there is simply too much to display.

[chogan@ecs-node-1 ECS-CommunityEdition]$ step2
> Pinging Management API Endpoint until ready
> Pinging endpoint (CTRL-C to break)

Complete setup via UI

Now we are ready to login to the UI, and finish off the checklist. Point a browser to https://<install-node> and you should get the ECS login as shown below. The default login credentials are root/ChangeMe.

You will have to provide a new password initially. Once you have created a new password, logged out and logged back in again, you can finish the install checklist by clicking on ‘Resume Checklist’:
The only outstanding task we have is to create our first S3 bucket. Just click on the ‘Create Bucket’ button and this will take you to the appropriate point in the UI.
I am now placed in Manage > Buckets. I am going to use the default user “object_user1” as the owner of the bucket. All I have to do is give it a name. I used “object_user_1_bucket” for the bucket name:
Now because this is the default user, it doesn’t have a suitable password at the moment, so that is the other thing I need to change. It’s very simple to do from the Manage > Users view. As you can see, the current password is ChangeMeXXXX. I can Delete this password, and create a new password by clicking on the Generate and Add Password.
With the new password created, I can use any S3 client to connect to the S3 object storage and access my bucket. I have been using an S3 Browser from NetSDK LLC to create, upload or read from the S3 buckets. You can use any number of clients for this task. Simply add the user as the Access Key ID (object_user1) and the secret key is the password we just generated:
And now I can access my bucket on the ECS:


This did take a little bit of planning to get up and running. However it does offer way more data services than just S3, which is the only one I have looked at so far. I’ll try to find time to investigate how well the other ECS services work on top of vSAN. Overall, a nice solution.

Exit mobile version