FreeNAS 10 and telemetry

[Ericloewe]

So I've been thinking about what could be broadly described as telemetry. Specifically, a way for FreeNAS 10 to send regular status updates via an interface to a device that's doing something with that data.

I think an example would help clear things up:

Imagine a hardware fan controller (since Supermicro's fan control is so simple and inflexible). It could receive regular updates of current disk temperatures (from SMART), CPU/System temperatures (from IPMI or directly from system sensors) as well as some load statistics (from the current reporting infrastructure) for some proactive fan control. In this case, the interface could be Serial (those RS-232 ports are of no use idle) or perhaps over IP networking.

Another scenario that comes to mind is a centralized monitoring system - for this one, IP networking sounds like the best option.

I wanted to get a feeling for what the community thinks about

• The general idea
• Interesting scenarios
• Data sources
  
• Interfaces

If this yields interesting results, I'll write up a feature request on the bug tracker.

 

[DrKK]

Interesting. But you know, at some point, https://pbs.twimg.com/media/CIXUpVlWoAIK3ta.jpg

I'd exercise caution adding more features like this. We should be pretty conservative with it.

 

[flatterlight]

Hi,
with the now (in FN 9.3) implemented SNMP we can get a lot of data for telemetry.
Currently you can get:

• Hardware and Software type
  
• CPU usage
• memory and swap usage
  
• networkinterfaces with utilisation and traffic type
• utilisation of various mountpoints
• Disk I/O

Flatterlight

 

[DrKK]

Yeah, I was going to say, I think we already provide most of this data via SNMP if you enable it...

 

[Ericloewe]

Yeah, I was going to say, I think we already provide most of this data via SNMP if you enable it...

Hi,
with the now (in FN 9.3) implemented SNMP we can get a lot of data for telemetry.
Currently you can get:

• Hardware and Software type
  
• CPU usage
• memory and swap usage
  
• networkinterfaces with utilisation and traffic type
• utilisation of various mountpoints
• Disk I/O

Flatterlight

Yeah, I had that general idea. For my particular potential use case, for instance, the temperatures would be missing.

The big point here, as I perceive it, are the sources of information. I'd like to hear what other stuff people would like to have.

 

[cyberjock]

This seems so overwhelming in terms of getting support for a wide range of hardware, etc that I don't see how you could realistically implement it for a project as small as FreeNAS with the limited developer resources we have. This really sounds like something that should be created by the FreeBSD project, then implemented in FreeNAS.

If you've built your server from server parts, you shouldn't have a need for this. And the amount of resources you'd need to make this work would be extraordinary. I, personally, would rather the devs work on stuff that really matters to a lot of people instead of some kind of unique proprietary monitoring system.

 

[Ericloewe]

This seems so overwhelming in terms of getting support for a wide range of hardware, etc that I don't see how you could realistically implement it for a project as small as FreeNAS with the limited developer resources we have. This really sounds like something that should be created by the FreeBSD project, then implemented in FreeNAS.

If you've built your server from server parts, you shouldn't have a need for this. And the amount of resources you'd need to make this work would be extraordinary. I, personally, would rather the devs work on stuff that really matters to a lot of people instead of some kind of unique proprietary monitoring system.

It would need to be kept simple. At this point, I think most can be done as an extension of the current SNMP stuff (for the transmission) and more coherent use of existing data (SMART drive temperatures are already processed, as are system temperatures and anything currently monitored but not exposed beyond the GUI).

If you've built your server from server parts, you shouldn't have a need for this.

No, fortunately, everything works well enough. It's mostly one of those "gee, I wish I could try something cool, but the data isn't easily grabbed from the server" moments. Nice to have, but definitely not urgent in any way.

 

[Glorious1]

You guys know far better than I do, but the exact example Ericloewe mentioned resonates with me. My Supermicro board allows limited control of my very nice fans. I can't really find a happy medium between fans going too fast 24-7 and good cooling when the room gets warm and you're doing long smart tests or scrubbing. When idle, a lower setting is fine, when busy/warm, it really needs a higher setting, even maximum. In my dream world, a fan controller supplied with all the necessary data (especially disk and board temps) would be pretty cool.

 

[Crispin]

I would be interested in this and happy to work with someone for it.
I am busy building a screen to mount in the 5.25 bay to display the basics. Current thinking is a two line LCD with an Arduino backpack to it.
I want to use it on my HP microserver and will either read from http (plugin in a jail) and just parse the info. v0.01 would be use scrolling info streamed to it. v2 would have menu driven buttons.