3 Nodes with 2 Drives each

[varet80]

Good day,
I am trying to migrate away from another solution. That solution supports 6 drives distributed in 3 nodes (2 drives each) with an Erasure Code 2 (if I am not stating it in a wrong way).

I use K8s and moving to RustFS I would like to keep the multiplication of my data to times 3 (original and 2 copies) in these 6 drives.

My option would be to have

• 6 Pods with 1 drive each (2 pods will run on the same node)
• 3 pods with 2 drives each

the 6 pod setup could use some kind of DC or similar parameter, whcih will help Rust identify the location of the data (think of Elasticearch Location and replicas).
OR I could have an EC setup in a way to support the second scenario, even if I am sacrifising 3 times the space.

Is there a way to achieve this setup, in order not to change hardware?

What I would like to achieve is: Having each object copied on each physical node at least once.

Thank you

[loverustfs]

Hi @varet80,

3 nodes, each node has two disks, a total of 6 disks.

The minimum number of stripes is 4, so it can only be divided into one stripe.

In a stripe, the M volume can only be set to 50% of the stripe at most. That is, EC:3.

Therefore, I think there is no way to meet your needs.

[varet80]

Can I meet the EC:3+3 then?

[loverustfs]

Can I meet the EC:3+3 then?

You can satisfy 3 + 3 stripes.
3 data disks, 3 parity disks.

Parameters in RUSTFS:
#Set stripes
RUSTFS_ERASURE_SET_DRIVE_COUNT=6

#Set check disks
RUSTFS_STORAGE_CLASS_STANDARD =EC:3

If you encounter any problems, we are happy to help you!

[varet80]

This means I can lose easily a Node with two disks! and still able to access my data.

[loverustfs]

This means I can lose easily a Node with two disks! and still able to access my data.

Yes,if you have any questions, please feel free to ask us.

[loverustfs]

If you need to implement such a function.

You should use bucket replication:
Write a copy of the data in one node and copy it to node2 or node3.

[loverustfs]

As with MinIO, once a stripe is set, it cannot be modified.

Therefore, if there is data, you need to back it up first.

[loverustfs]

I want to setup my system from scratch and then move data from MinIO. Migrate away from it.

Thank you very much!

If you encounter any problems while using this app, we'll be happy to help!

[varet80]

I thank you for the help till now!

A background. This is a Home setup, but I run k8s (talos) at Home and in DC for my personal stuff.

What Object store is for:

• Nextcloud with backend of S3
• Thanos
• Photo Galery
• Different device backups

and I always mirror these to a remote Object store.

Thus the limitation of nodes, as Private individual I try to seek for balance between availability and cost.
Soon might have more questions!

[varet80]

@loverustfs One question,
I am looking for a concentrated documentation of all possible parameters (Spec) of Operator CRD
And/or the Cofig.toml with all possible params.

As I cannot easily find it? Seems I am not good on searching.

[loverustfs]

@loverustfs One question, I am looking for a concentrated documentation of all possible parameters (Spec) of Operator CRD And/or the Cofig.toml with all possible params.

As I cannot easily find it? Seems I am not good on searching.

Hi @varet80 ,

Our current support for Kubernetes is still insufficient.

This includes Helm, YAML, and Operator. We are currently developing a new project: rustfs-operator. https://github.com/rustfs/operator

We hope this project can alleviate the difficulties of Kubernetes deployment.

[loverustfs]

I thank you for the help till now!

A background. This is a Home setup, but I run k8s (talos) at Home and in DC for my personal stuff.

What Object store is for:

• Nextcloud with backend of S3
• Thanos
• Photo Galery
• Different device backups

and I always mirror these to a remote Object store.

Thus the limitation of nodes, as Private individual I try to seek for balance between availability and cost. Soon might have more questions!

You're most welcome!

If you have any specific questions, we'll do our best to help you.

Thank you for your trust in open source and in us!