docs: provide hardware recommendations for reference architectures #12534
Conversation
|
we may want to link/add to the architecture diagrams i've added here: #12584 |
|
|
||
| | Users | Node capacity | Replicas | Storage | GCP | AWS | Azure | | ||
| | ----------- | -------------------- | -------- | ------- | ------------------- | -------------- | ----------------- | | ||
| | Up to 2,000 | 4 vCPU, 16 GB memory | 1 | 1 TB | `db-custom-4-15360` | `db.t3.xlarge` | `Standard_D4s_v3` | |
There was a problem hiding this comment.
We've been using a db-custom-8-32768 for our 2000 user scaletests, so these CPU number here may be slightly inaccurate. Granted, for a regular (non-scaletest) deployment, I think 4 vCPU is borderline sufficient as it's the 2000 active workspaces that push the DB CPU load up to ~80% (8 vCPU).
There was a problem hiding this comment.
On the other hand, our scaletesting methology was really aggressive, so I would expect a similar pattern on the user side 🤔
What is your recommendation @maf? Should we switch to db-custom-8-32768?
|
|
||
| When determining scaling requirements, consider the following factors: | ||
|
|
||
| - `1 vCPU x 2 GB memory x 250 users`: A reasonable formula to determine resource |
There was a problem hiding this comment.
I think this better matches our reference arch?
| - `1 vCPU x 2 GB memory x 250 users`: A reasonable formula to determine resource | |
| - `0.5 vCPU x 2 GB memory x 250 users`: A reasonable formula to determine resource |
There was a problem hiding this comment.
Thinking about the future, I would leave 1 vCPU, WDYT?
There was a problem hiding this comment.
We can probably argue for wiggle room here based on how certain Terraform providers may be more CPU-intensive than others.
|
Thanks for the reviews, folks. I'm going to merge it and we can implement the next changes in follow-ups. |
Related: #12426
This PR describes hardware recommendations for Coder references architectures including node sizing for
coderd, workspace, provisioners, and the database.@cian @mafredri
I reviewed the latest state of the art for scale tests, Grafana dashboards, and the current version of Scaling Coder. Feel free to adjust the numbers to be more accurate. I admit that I tried to give extra CPU/mem capacity.