Skip to content

Deployments requirements and proposal - #140

Draft
kgodey wants to merge 7 commits into
masterfrom
deployments
Draft

kgodey wants to merge 7 commits into
masterfrom
deployments

Conversation

@kgodey

@kgodey kgodey commented Jun 26, 2025

Copy link
Copy Markdown
Contributor

No description provided.

@kgodey kgodey changed the title Deployments requiremeents and text Deployments requirements and proposal Jun 26, 2025
Base automatically changed from templates to master June 30, 2025 08:00
@zackkrida zackkrida assigned zackkrida and unassigned zackkrida Jun 30, 2025
@zackkrida
zackkrida requested a review from mathemancer June 30, 2025 19:04
@zackkrida zackkrida added the pr-status: review A PR awaiting review label Jun 30, 2025
@zackkrida
zackkrida self-requested a review June 30, 2025 19:04
@zackkrida zackkrida self-assigned this Jun 30, 2025
@zackkrida
zackkrida marked this pull request as ready for review June 30, 2025 19:05

@mathemancer mathemancer left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

TL;DR

After giving a shot at reviewing this in depth, I don't think the project framework (as we've set it up) is appropriate for this work. The requirements doc isn't scoped, and probably can't be. Moreover, the actual work is a bunch of marginally-connected projects (as noted in your work). There is no such thing as "done".

I think it would make sense to reframe this as a set of experiments we want to do, put them in order, and then just do them on an experimental basis in that order.

I spent quite a bit of time trying to go through this and make it fit into (my understanding) of a "project proposal" and "project requirements" framework. I ended up running out of steam, but I can continue the review later if you think it's valuable, @kgodey .

Review

I have some line-by-line notes, as well as some in this main review (for older stuff that wasn't in the diff so couldn't be assessed line-by-line in the GH interface). It's not clear to me that we should spend the effort to try to improve this "project" to fit the framework, for the reasons I stated above.

Additional deployment options are actually solving four different problems:

  1. Makes it easier for people to evaluate Mathesar quickly.
  2. It makes it easy for people with existing infrastructure to integrate Mathesar into their infrastructure.
  3. It helps new people discover Mathesar (e.g. Supabase users looking at their marketplace, or Vercel users trying to integrate a DB)
  4. It makes it easier for people to actually deploy Mathesar in production.

This document does not make it possible to prioritize between these goals.

I thought the "How it helps users" section was hard to get actionable information from.

Offering more deployment options is relatively low-effort and high-impact.

Depends on the deployment options, I suppose. We could easily sink a month of developer time into "helm charts" since that's open-ended and highly dependent on the goals of the person using the charts. It's not at all clear to me that we have the information necessary to write ones that would be useful except for internally. (Perhaps we should start by trying to write a helm chart to stand up demo infrastructure automatically?).

This does not add any useful information, and the linked issues do not provide much useful information either. For example, the "I want an AMI" issue linked ended up being a problem that wouldn't have been solved by providing an AMI.

  • People with existing infrastructure are having to spend a lot of time and effort into integrating Mathesar into their workflow.

More detail needed. Who? What did they spend time on? Why would more deployment options help them? Which deployment options? How can they be prioritized? etc.

Use Case commentary

Alex

Why does he need instructions for TLS setup? He already knows he wants to use Nginx and Certbot. What instructions are needed for someone who already has that knowledge base?

Blair

This one seems good to me.

To be continued (maybe)

Comment on lines +133 to +135
Brent & Anish to implement, Kriti reviews user-facing or admin-facing changes as they are merged.

**Rough timeline:** 1 type completed per week.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe our estimations of the time involved are different. If you want to ship one of these per week, it would need to be the main thing someone is working on that week. Maybe that's the intent? If so, I don't think it's a good idea to make this backend-exclusive.

For example, the GCP one is going to be multiple days of discovery, tinkering, and then writing and reviewing documentation. I don't expect the turnaround on review (much less discovery and any implementation) to be a single week.

Comment on lines +122 to +124
* Trying to make our documentation too preemptively perfect – let’s just put it out there.
* Not implementing things because of prioritizing hypothetical maintainability concerns, or not imagining that something might be useful, or worrying about cluttering the documentation (these have all come up in the past).
* Spending a lot of time stuck on one installation method without raising it / figuring out if we should keep going.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It was at this point that I realized we should probably move this work into the "experimental" framework.

  • It's completely open-ended
  • We don't really have any information to assess the value of different platforms, etc. That's okay (if we're making sure we keep all changes reversible)
  • It's not actually a cohesive project. It doesn't have an end.

Comment on lines +158 to +190
### NocoDB

For what it’s worth, NocoDB supports a few different types of hosting platforms, see all types below in the next section.

**“Auto upstall”:** This is their default script meant for a Linux server that installs everything including Postgres, sets up a DB, sets up SSL, upgrades, etc. I think it’s meant to be idempotentent.

**Simple platforms:** Digital Ocean, Railway, some other “one click deploy” buttons.

**Cloud marketplaces:** AWS Fargate, GCP Cloud Run

**DevOps tools:** Docker, Docker Compose, Helm chart

**Local installation:** Homebrew, Nix/NixOS, npm

**Self-hosted PaaS tools:** Cloudron, CapRover \+ they link to an external TrueNAS / FreeNAS guide.

### Baserow

Baserow supports a similar list of hosting services as NocoDB.

**DevOps tools:** Docker, Docker Compose, Helm chart, K8S. They have images that contain all dependencies and are ready to go.

**Simple platforms:** Digital Ocean, Heroku, Render, Railway \+ a page for other third-party hosting providers

**Cloud marketplaces:** AWS Fargate / other ways of deploying on AWS using Docker & Kubernetes

**Self-hosted PaaS tools:** Cloudron

Additionally, they have:

- Guides on how to set up various proxies: Traefix, nginx, Apache
- A guide on how to set up logging and monitoring
- A guide on how to save files securely

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

To be frank, this seems like the most relevant part. It should be emphasized, and we should probably just list the union of the deployment options they provide, and prioritize that list.

@zackkrida zackkrida left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@mathemancer I see where you're coming from in the thinking that this shouldn't be a project. However, I think it could be if we add a few more specifics, such as criteria for when we could consider this work "complete".

@kgodey Brent's comments are extensive so I'll let them be the primary focus of your re-review. I am comfortable moving forward with this as a project after some refinement around more specific goals, or moving it into discrete experimental issues for each provider we want to consider adding.

Comment on lines +10 to +12
## Solution

### Supabase

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Somewhere in these documents I think we need a high-level overview of the solution. Between these two documents we never actually state something like "We will document multiple methods for deploying Mathesar in our documentation on docs.mathesar.org"

@zackkrida
zackkrida marked this pull request as draft August 18, 2025 21:03

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

pr-status: review A PR awaiting review

Projects

No open projects

Development

Successfully merging this pull request may close these issues.

3 participants