Skip to content

healthcheck: render major deployment health issues in dashboard #7748

Closed

Description

@bpmct

Would be awesome if the product could immediately tell users if there is a severe issue with their deployment, and how to fix it. For example, an unreachable access URL, DERP errors, WebSockets are disabled, etc. This will require collaboration with @bpmct, @kylecarbs, or @ericpaulsen who can help with wording and writing docs around how to resolve these issues.

It feels like we should show this as a service banner on all pages since pass: false indicates a major issue. The goal of this is to allow users to self-diagnose issues quickly instead of them having to spend hours debugging themselves and then contacting us for help (where we may tell them to navigate to a specific page or endpoint to reveal errors).

Acceptance criteria

Avoiding false positives: A non-dismissable banner may be frustrating if the deployment is, in fact, healthy, despite a failed health check. If we can think of common cases where this would happen, we should ideally fix our health check logic. However, I don't think this needs a ton of testing. If necessary, we could always revert or make the warning dismissable.

Performance: We should run basic performance tests or make design decisions to ensure this does not significantly slow down page loads.

Visibility: Since this banner will only show up during major issues, I don't have strong opinions on whether users should see the banner or only admins. I'm not super concerned about the banner disrupting UX since Coder would be unusable at that point and users seeing it may lead to a quicker resolution time from an admin.

CLI: I don't think this necessarily has to show up as CLI errors, but it would be nice (e.g. similar to a warning that Coder is out of date)


To avoid doing a bunch of calculations on the front, this feels like its partially blocked by #7745 (top-level fail reason)

Activity

  1. mtojek commented on Jun 1, 2023

    @mtojek
    Member

    Nice idea! I'm curious if instead of a dashboard it is better to provide Coder Status Page, so people can just navigate there. Something like: https://health.dev.coder.com or https://dev.coder.com/health.

    We may have some background jobs running to examine the state of deployment, and depending on the user role we can give more or less detailed insight into the situation.

  2. bpmct commented on Jun 1, 2023

    @bpmct
    MemberAuthor

    For this issue, I was thinking of just putting all major errors as a service banner on the top of every page to make the page (and links to resolve/troubleshoot) as obvious as possible.

    One concern I have about starting with a separate health page is that it may not be immediately obvious, particularly for a new admin, that they can navigate to that URL to get info about the issue they are running into. They may still spend hours troubleshooting, finally contact support, then learn about the health check page.

    I am definitely in favor of a health page with more detailed info down the road. For example (your networking is over websockets, which is 25% slower) 👍🏼

    Also, if you have ideas on how to make the health page discoverable, particularly when there is a major issue (e.g. an issue that would prevent any user from connecting via SSH), then I am fine doing that.

  3. added this to the ❓Sprint 1 milestone on Jun 12, 2023
  4. added
    must-doIssues that must be completed by the end of the Sprint. Or else. Only humans may set this.
    on Jul 27, 2023
  5. BrunoQuaresma commented on Jul 31, 2023

    @BrunoQuaresma
    Contributor

    I'm not sure how to present this to the user so I will create a design first.

  6. mtojek commented on Aug 8, 2023

    @mtojek
    Member

    @BrunoQuaresma Is this issue still blocked?

  7. BrunoQuaresma commented on Aug 8, 2023

    @BrunoQuaresma
    Contributor

    I put this in blocked just to let more visible we only want to close this when the health page is GA but IMO, we should close it and file new tickets for the improvements and for the "remove out experimental" work if we need to. cc.: @bpmct

  8. bpmct commented on Aug 8, 2023

    @bpmct
    MemberAuthor

    Closing in favor of #8971

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

must-doIssues that must be completed by the end of the Sprint. Or else. Only humans may set this.

Type

No type

Projects

No projects

    Milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions