Rancher Performance Monitoring

I have 18 containers running on my host. What’s I’d REALLY like is if the general health of my host could be communicated to me on the main splash screen. For instance, how close is the host to consuming all of it’s HD or memory? What’s the biggest consumer of memory from a container perspective. The problem gets even worse when you start adding hosts - you don’t have the time to click into every container to see their resource utilization. I’d love a “one stop shop” view for system health.

Also, it would be awesome if you could configure email alerts for performance stats. For instance, I’d like to know if ANY of my hosts reach over 90% memory utilization. Or, if any of my containers use more than XXX of the filesystem.

Thoughts?

We are adding more stat aggregations to our UI where you can click on things like Environments, Hosts, Service, Stacks, etc. to get more information on cpu/mem/disk utilization rather than clicking one container or host at a time. This will probably be out in a couple of weeks.

In terms of alerts, we are looking into generating more events within Rancher and provide you with a way to subscribe to it.

Would be nice if those stats could be sent to a 3rd party system, without too much hassle… for us who already have the infrastructure for it with alerting etc.

@kaos With the current version of rancher-server you can already stream stats for a host and container to any client.

Go to the API page for a Host

The stats API for a host/container can be found under the ‘links’ section of the API page for host/container.

Open the stats API, you’ll find an URL and a token.

You can make a Websocket request to the URL with the token as a query param.The long lived Websocket will stream stats as long as the connection is alive.

The same steps can be used to retrieve container stats as well.

In the upcoming releases, we’ll be adding stats links to Environments, Services, and Stacks in addition to Hosts and Containers.

2 Likes

Ah, that is really cool, thanks.

I did click around a bit on the api pages, but didn’t really figure them out yet. Should’ve guessed that I could access the stats from there (I have a hunch that the GUI is using the very same api for all its functionality, and given the UI shows stats, I could’ve realized this on my own :stuck_out_tongue: ).

Cudos to @ibuildthecloud et al for the way the api is exposed. Exemplary :+1:

There is some work to do to make it more usable but yes, the UI is 100% driven by the API so you can do anything it does (and some stuff it doesn’t yet).