Cloudformation deploy is recommended for production?

Hi!

The easiest way to start on AWS seems to be the CloudFormation stack provided by MemSQL.

Is that a recommended way to have a production cluster?

I’m assuming that the AMI used and default settings are optimised with the best practices for AWS as per the documentation. Is that true?

Operations like adding/removing hosts (leafs/master/etc), performing instance class changes, and so on, should be performed changing the CF stack or manually?

Thanks,
Vinicius

Hi @vkruoso

We highly encourage you to run MemSQL via CloudFormation in production. Here’s the install guide. There is also a section with Advanced Installation instructions that will answer your questions.

Thanks!

Hi Vinicius,
Let me add some more information to what Jacky wrote above. That path is via our docs.memsql.com site.

We also have MemSQL listed as a product in AWS Marketplace. Link here https://aws.amazon.com/marketplace/search/results?x=24&y=14&searchTerms=MemSQL
We have a BYOL and a PAID version. For the PAID version you can pay by the hour for MemSQL usage.
The BYOL option allows you to create very large clusters up to 100 nodes.
The PAID listing allows you to create clusters up to 10 nodes.

The CloudFormation Template (CFT) used in both the AWS as well as our docs link is the same.
The AMI in all the above CFTs are optimized to use OS configuration parameters appropriate for MemSQL and it also uses the faster ‘gp2’ Volume type for the EBS.

Currently we do not support cluster expansion or shrinking with the CFTs mentioned above.

Very soon, we are planning to release an offering which will allow true elasticity in terms of on-demand cluster expansion. Stay tuned!

Thanks for your interest,
Ramesh

Hi! @Jacky and @rnarayanan,
Thanks for your answers.

The CFT seems very limited in the sense that we have very little customisation options. We will be running MemSQL on a budget, so having 1TB disks on all nodes just adds a cost that is not necessary in our use case, for instance. Also all operations like backups, disk resizing, and new nodes configurations must be made manually.

Looking forward to test the on-demand offering.

Thanks,
Vinicius

Hi Vinicius
Many of our users download the template from the sites mentioned - either via docs or AWS, modify it and use them. The stack creation process allows for uploading your modified template.
For disk size, you just have to change the VolumeSize parameter (it is 1024 now).
Yes, backups have to be handled manually but that is true even for on-prem installations.

If you are able to estimate your size requirements in the beginning, then there is not really a reason to change the cluster configuration. Let us know if you have any related questions on that.

Thanks,
Ramesh