Implement container health alerts

[Jdbarad]

Container Health Alerts & UI Overhaul [#716]

📃 Description

This PR introduces comprehensive support for container health alerts and significantly overhauls the Alerts UI. It adds backend logic to monitor container health status and a redesigned, searchable frontend interface for efficient alert management in environments with high container counts.

📖 Documentation

Add a link to the PR for documentation changes.

🪵 Changelog

➕ Added

• Container Health Alerts: Backend logic to monitor and trigger alerts based on container health status.
• Searchable Container List: A new UI section in the alerts sheet to quickly find and toggle alerts for specific containers.
• Database Migration: A dedicated migration to convert the alerts collection name field to text, supporting dynamic container alert names.
• Batch Actions: Added "All" and "None" buttons to quickly toggle alerts for all currently filtered containers.

✏️ Changed

• Alerts UI Overhaul: Refined the layout of the Alerts sheet, removing excess padding and implementing a focused scroll area for the container list.
• Stats Serialization: Exported Health and Status fields in container stats for frontend and alert system access.
• Alert Logic: Updated the alert manager to handle container-specific thresholds and notification subjects.

🔧 Fixed

• UI Consistency: Standardized container alerts to use the same toggle and slider components as system-level alerts.
• Reliability: Improved container data fetching using useEffect to ensure the list stays in sync when switching between systems.

🗑️ Removed

• Unused Code: Cleaned up various unused imports and state variables in the alerts components.

📷 Screenshots

[albertorizzi]

Is it possibile to set for every container the trigger of unhealthy instead to set for everyone?

[kaysond]

Is it possibile to set for every container the trigger of unhealthy instead to set for everyone?

Agree. The ux is not great IMO. The original issue had a much nicer mock up:
#716 (comment)

[Jdbarad]

Is it possibile to set for every container the trigger of unhealthy instead to set for everyone?

Agree. The ux is not great IMO. The original issue had a much nicer mock up: #716 (comment)

Sure @kaysond

[Jdbarad]

@albertorizzi @kaysond : Check this approach. If it's still not functional, then I can try some different approach.

[svenvg93]

Just my two cents on the UI;

• Keep the All Containers in this sheet.
• Move the settings per container to the Container Table on the System page. Same way as we have it for the System alerts, with a little bell opening a sheet with the alerts for that container. It will allow us later to add other alerts as CPU, Memory etc per container as well.

[Jdbarad]

Just my two cents on the UI;

• Keep the All Containers in this sheet.
• Move the settings per container to the Container Table on the System page. Same way as we have it for the System alerts, with a little bell opening a sheet with the alerts for that container. It will allow us later to add other alerts as CPU, Memory etc per container as well.

Agreed. Keeping the sheet focused on 'All Containers' and shifting the specific alerts to the Container Table is a cleaner approach. I’ll rework the UI now and get back to you with a fresh, updated look.

[albertorizzi]

Any news?

[henrygd]

Sorry for lack of guidance, this is a fairly complex feature and I don't have enough time to go super in-depth right now.

All current alerts are based on systems. Rather than modifying the current alerts collection, it may be easier to create a new container_alerts collection. Likewise, it would probably be cleaner to make a separate alerts_container.go rather than mixing this in with alerts_system.go.

The functionality should work more like alerts_status.go anyway, since container health is a boolean state (unhealthy or not unhealthy), like system status (down or not down), rather than a value we need to average out and compare to a threshold like in alerts_system.go.

I'm also not completely satisfied with what I did in alerts_status.go, so it's not necessary to copy that pattern exactly. I'm sure there's a better way to do it.

We have access to all the container data on the hub side in system.go, so we can pass the data through to the alert functions somewhere around here:

beszel/internal/hub/systems/system.go

Lines 192 to 198 in afc19eb

	// add containers and container_stats records
	if len(data.Containers) > 0 {
	if data.Containers[0].Id != "" {
	if err := createContainerRecords(txApp, data.Containers, sys.Id); err != nil {
	return err
	}
	}

IMO it would be best to make it a general "Container Health" alert that sends notifications when any containers on the system switch to / from unhealthy. We can add a way to exclude certain containers later, it doesn't need to be included in this PR. I'd rather keep the scope of this one as narrow as possible.

[Jdbarad]

@henrygd Soon, I'll share the architecture change and UI in detail. In my mind also same thing is going on about container_alerts collection. Once all and you review i'll start working on that part.

[Jdbarad]

@henrygd @albertorizzi Check this UI screenshots

[albertorizzi]

@Jdbarad I use Notifiche Webhook / Push with Telegram. Is this flow implemented?

[Jdbarad]

@Jdbarad I use Notifiche Webhook / Push with Telegram. Is this flow implemented?

Yes, That flow also implemented

[Rubilmax]

Pls review this devs 🙏

[Jdbarad]

@henrygd Can you please review this updated code.

[ddmunhoz]

I quickly reviewed it, everything seems to check. 🚀

[henrygd]

I should be able to get to it this week, or at least relatively soon. Don't worry about merge conflicts.

[Jdbarad]

@henrygd Let me know if any help is needed.