Add support for push-based health checks

[unaizorrilla]

Hi,

Saving/tracking the whole history of health check information is very usefull on many enterprise system.

Moving health status and other related information as metrics to Application Insights , Promethus etc can be a awesome feature to be included on Health Checks.

There is any plan to include this? PR can be accepted?

Regards Unai

[mkArtakMSFT]

Thanks for contacting us, @unaizorrilla. @glennc, is this something we plan to add support for?

[rynowak]

Hi @unaizorrilla - we planned for push based health checks to work by registering a long running service that calls into the health checks service at regular intervals. I think this will work better since you can control the interval, and it doesn't rely on traffic to the health endpoint to function correctly. Also it can be used by apps that aren't doing http at all!

We think that we don't really need to put a feature in the box for this, and I was going to write a sample.

What do you think about this? Have I understood what you're asking for? Should I make the sample and then we can discuss further?

[unaizorrilla]

Hi @rynowak!

We don't talk about the same topic :-) my fault!!

My feature request is not about executing IHealthCheckService periodically with IHostedService, configured liveness / readiness on K8S,availability alerts on AppInsights or something similar.

The "trackers" idea is the same i do on BeatPulse:

https://github.com/xabaril/beatpulse#tracking-pulses

With trackers you can send the HealthCheckReport information to third parties ( Application Insights, Prometheus, etc.) This information is very usefull to create dashboards and KPI with availability service metrics or dependencies response times.

The design for this request can be:

• the abstraction for define health check tracker

public interface IHealthCheckTracker { Task Track(HealthCheckReport report) }

• the availability to register it on IHealthCheckBuilder or directly on the service provider.
• Modify IHealthCheckService to send HealthCheckReports on all registered trackers…

Regards Unai

[rynowak]

OK, it appears we understand each other but I'm still missing something.

• An IHostedService implementation would check health regularly and push the results somewhere external
• A 'Tracker' implementation would listen to health reports being produced and push the results somewhere external

Why is it better to have the latter for feeding data to app insights/prometheus?

If your goal is to capture detailed for correlation and offline analysis then we already feed that into logging. Is this not good enough?

If your goal is try and use health checks as a metric system, wouldn't you want that to be independent of someone visiting the health endpoint?

[unaizorrilla]

"If your goal is try and use health checks as a metric system?"

Yes

"wouldn't you want that to be independent of someone visiting the health endpoint?"

• Well, because at this moment the /[health] path only respond with status and the metrics can be more detailed with response times, dependencies ( sql server, redis .. ). Of course is posible create a new ResponseWriter but this probably wouldn't be necessary for reqular health check request.

• Because is more simple if you don't need to build nothing more to get this metrics available. AddApplicationInsightTracker and all is working!

Thanks again! Unai

[mkArtakMSFT]

Thanks for contacting us. We believe that the question you've raised have been answered. If you still feel a need to continue the discussion, feel free to reopen it and add your comments.

[rynowak]

We're implementing push-based health checks so that @unaizorrilla can port his code on top of it. This item tracks that work.