Stories
Slash Boxes
Comments
Search

News for nerds, stuff that matters

Slashdot Log In

Log In

Create Account  |  Retrieve Password

OpenBSD 3.9 Adds Sensor Framework

Posted by ScuttleMonkey on Mon Mar 27, 2006 06:21 AM
from the now-even-my-computers-are-calling-me dept.
Operating Systems Software BSD
wbglinks writes to tell us ZDNet is reporting that the newest version of OpenBSD will include a sensor framework to help system administrators keep tabs on the environmental conditions of their servers. From the article: "At present, there are a number of commercial products that allow the environmental conditions of servers to be monitored, but different brands of server require different products. For example, Dell PowerEdge servers use the Embedded Server Management tool, while Sun Fire Servers use Sun's Remote System Control. This can make server management tricky when running a heterogeneous architecture. OpenBSD 3.9, which is scheduled for release on 1 May, includes support for the sensors and the sensor management tools used on a number of architectures."
+ -
story

Related Stories

[+] BSD: OpenBSD Project in Financial Danger 610 comments
DieNadel writes "In an entry to the OpenBSD Journal, Marco, from the OpenBSD project, warns about the somewhat disturbing financial situation in which they are now. The OpenBSD team is the one that also develops the OpenSSH suite, used nowadays almost everywhere. From the entry: 'What I want to point out what a lot of people don't seem to realize is that OpenSSH development is paid from the same pool of money as OpenBSD. OpenSSH is in use by millions around the world however the revenue stream just simply isn't there. This is where other projects could help. Without naming entities or projects by name there are others out there that are sitting on some cash. It would be wonderful if these entities could share some of the wealth to keep us going.'"
[+] BSD: Theo de Raadt Discusses OpenBSD and Beyond 476 comments
emil writes to tell us that NewsForge (Slashdot Sister Site) is running an interview with OpenBSD project leader Theo de Raadt. In the interview Theo explores the upcoming release of OpenBSD 3.9, continuing financial difficulties, and some of the tension between the OpenBSD team and other businesses that some feel are taking advantage of the free software without giving anything back. In related news the Jem Report has an interesting writeup that expounds on widespread difficulties that could be faced if the OpenBSD project continues its downward spiral because of their parallel development of OpenSSH.
This discussion has been archived. No new comments can be posted.
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.

OpenBSD 3.9 Adds Sensor Framework 50 Comments More /

 Full
 Abbreviated
 Hidden
More
Loading... please wait.

  • sensors and slashdotting (Score:4, Funny)

    by cabinetsoft (923481) on Monday March 27 2006, @06:28AM (#15001674)
    De Raadt has already been using the sensor framework to monitor the machines running in the project's server room. "I now get a call on my cell phone whenever something is wrong in the machine room," he said.
    and I bet the temperature warning reads something along the lines of "Link to your site posted on slashdot.org"

  • is_computer_on_fire() (Score:1, Funny)

    by Anonymous Coward
    while Sun Fire Servers


    Finally some use for BeOS' is_computer_on_fire() [tycomsystems.com] function!
  • what's the situation in Linux? Is this the same thing as the 'hardware sensors' modules in the kernel?
      • However it will most likely not be truly cross platform. I will be waiting for the Windows versions very eagerly. When you start talking about heterogenous environments, there will be Windows servers around. You can't get around that no matter how zealot and OSS fundamentalist you are. Mostly likely it will portable - like most of OpenBSD's code. Just don't expect them to port it for you. With some effort, probably it would work on unix services for windows (which is based on openbsd btw).

  • Which means... (Score:4, Insightful)

    by MadMirko (231667) on Monday March 27 2006, @06:58AM (#15001756)
    ... they add support for BMC and IPMI?

    Which, while fine in itself, is hardly a groundbreaking achievment for an OS, or is it? At least Windows has done that for years, and I believe Linux does as well (at least we have a working "sensor" implementation on a few RedHat / HP servers).
    • We will also add support for this in INM if possible.
    • Linux has many hw/temp sensor monitors and programs (mbmon, lbmon, etc). However (like all major sub-systems in Linux), it does not have a framework to give you a single set of commands that work with all hw/temp sensors out there.

      If you want to monitor systemX, you use programX. If you want to monitor systemY, you use programY. And so on. (It's like the horrid mess that is ifconfig/iwconfig/wiconfig/athctl/younameitctl).

  • Welcome to.... (Score:3, Informative)

    by diegocgteleline.es (653730) on Monday March 27 2006, @07:19AM (#15001819)
    "There is a significant new sensor framework [in OpenBSD 3.9], which supports voltage sensors, fan sensors, temperature sensors, and so on," said de Raadt. "Such a feature is still missing in Linux and other major operating systems."

    There we go [kernel.org]
      • Theo wants it both ways, he proudly boasts "I don't know what other OSes do, I don't use them" until someone tells him that his feature isn't on Lunix and then it's in the press release.

        • "I don't know what other OSes do, I don't use them"

          (Disclaimer: I know Theo didn't said that himself)

          Well, then maybe he should just shut up about the things he doesn't know. Linux DOES have support hardware monitoring drivers, IPMI, and the randomization feature since 2.6.12 or so. Pretty much EVERYTHING he said about Linux is wrong.

          If you don't know nothing about something, then just don't talk about it. What Theo did is pure and true FUD.
        • I think you are confusing your attribution. The quote you are repeating is correctly attributed to Linus Torvalds, not Theo De Raadt.
          • Thanks for the link & the rant.

            I'm not knocking Theo or OpenBSD (my company's web server is 3.8) , just found it funny him using Linux as a benchmark when he doesn't use it.

            Linus is right though, there's nothing in XP or 5.x worth copying =)

            They should both use plan9 a bit more, then they'd know what was cutting edge.

      • Re:Welcome to.... (Score:5, Informative)

        by Triumph The Insult C (586706) on Monday March 27 2006, @10:31AM (#15003007) Homepage Journal
        this sensor framework is integrated into the base install. it is managed and developed by the openbsd developers, not a third party group where changes still have to get imported

        the framework supports a lot of sensors. along with sensorsd(9), it is a large improvement over what has been available for other OSes
        • You know, the Linux sensor developers ARE Linux developers (in charge of a given subsystem>), it's not a "third party group" where "changes still have to be imported" - in Linux the hardware monitoring features, IPMI etc have been in the mainline tree - and shipped in distros with commercial support etc - for years.

          I really don't see the difference, except that OpenBSD seems to be the one who is catching up.
      • I have not had such great luck with lm_sensors. I usually run IBM hardware (not a lot of choice in the matter), and it wasn't until pretty recently that you could get lm_sensors to work on a lot of it. This seems to have a lot to do with the corruption issues lm_sensors initially had with Thinkpads--the developers seem to have been (justifiably) skittish about IBM hardware. The versions of lm_sensors packaged with the distributions we use (enterprise versions of SUSE and Red Hat) have also been well behind
  • Have I been missing this section this whole time, or is this something new?
  • for me (my whole world is snmp, it seems) I'd want to know if there is any good progress on getting remote mgmt via snmp working better than it has, in the past.

    for example, sun has the 'platform mib' and 'entity mib' and in these two (as a sum) you can get voltage and fan speed and temperature and even alerts (traps) when thresholds are reached.

    I have not seen the entity mib (for example) on ANY lower end unix platform (freebsd, linux, etc). maybe I have to be the one to write one...

    getting sensor data ha
  • I thought Slashdot readers were opposed to sensorship. (*Rimshot*)
  • Wow, this latest move is sure to rocket OpenBSD to the top.

    I mean, the network performace will likely still suck [bulk.fefe.de], especially compared to the competition, but at least now we can monitor our servers!

    Big Brother's [bb4.org] given us this capability for years. Nothing to see here, move along.

  • Not NEW in the slightest... (Score:3, Insightful)

    by evilviper (135110) on Monday March 27 2006, @06:59PM (#15007302) Journal
    This "New Sensor Framework" has been in the mainline kernel since 3.5, and working quite well, thank you. I certainly wish other OSes would get this stuff built-in (of course OpenBSD is also lacking a lot of good features that FreeBSD/Linux DO have).

    Setting up lmsensors was an infuriating and disgusting mess on Linux. After an hour of kernel recompliations, and i2c/lmsensors version mis-matches, I just gave-up. I decided to simply parse the output of mbmon (most trivial setup, EVER!).

    • Well if you don't shell out money for shifts and stand by calls don't expect work to get done 24/7. Well at least in most european countries it works this way, don't know about the rest of the world.
        • Lucky you! Last job I was on call every 5th week, for 7 days, 24 hours a day and got a page (to which I had 20 minutes to respond) every 15 minutes or less. Which is why it was my previous job. The pay was nowhere near enough to make up for losing 7 days of sleep.
          And, as far as salary goes, funny how salaried US employees tend to work a lot of those 80 hour weeks, and very few of the 20 hour versions.

    • Re:File cabinets and fires (Score:4, Funny)

      by merdaccia (695940) on Monday March 27 2006, @07:06AM (#15001782)
      Remember when you could go back to work on Monday and find a disaster that would take you three weeks of painstaking work to fix because you had no way of knowing a fan died?

    • Remember when you could go home on the weekend and actually enjoy yourself? Sensor management means that you can never be completely away from your cellphone.

      Sensor management or no sensor management it's pretty the same thing... instead of the server dialing / paging you there can be a human dialing / paging you anyway. And of course YOU CAN switch off your cell phone if it's bothering...

      This reminds me of some time back when I used to tech support for a telco logging system. I was out with my frie

    • Sensor management means that you can never be completely away from your cellphone.

      Sensor management means that you will be aware of problems as they are in the nascent stages of development, before they become a crisis. It provides you the time needed to research and repair, instead of the panicked "fix it now!" when systems stop working.

    • Ehm, in that part of the interview he's talking about "randomised memory allocation", not about sensors.

    • Re:Should it be in? (Score:3, Informative)

      by Anonymous Coward
      If you RTFA, you can see that that quotation was taken out of context. Theo was discussing fully random memory allocation to prevent buffer overflow. As far as I know, sensor monitoring is available quite easily in Linux.
    • There's lots of niche features which are in the main branch of the kernel.

      NUMA, OMAP, powerPC, and the list goes on and on.

      However, I think it would be VERY cool to be able to query /dev/tempsensor1 for the tempature of my motherboard or CPU. Might even be able to do something useful with it.

      Cheers,
      Ben
      • cat /proc/acpi/thermalzone/THRM1/temperature
        • And that's the ACPI thing, you've a sysfs interface [kernel.org] for other sensors.
        • cat /proc/acpi/thermal_zone/THRM/temperature

          is the spot for me on an Gentoo amd64 laptop

          temperature: 48 C

          for those curious as to it's output)

          • $ uname -srm
            OpenBSD 3.6 i386
            $ sysctl -a hw.sensors
            hw.sensors.0=viaenv0, TSENS1, temp, 24.30 degC / 75.74 degF
            hw.sensors.1=viaenv0, TSENS2, temp, 61.40 degC / 142.52 degF
            hw.sensors.2=viaenv0, TSENS3, temp, 5.10 degC / 41.18 degF
            hw.sensors.3=viaenv0, FAN1, fanrpm, 0 RPM
            hw.sensors.4=viaenv0, FAN2, fanrpm, 0 RPM
            hw.sensors.5=viaenv0, VSENS1, volts_dc, 2.52 V
            hw.sensors.6=viaenv0, VSENS2, volts_dc, 2.43 V
            hw.sensors.7=viaenv0, Vcore, volts_dc, 1.98 V
            hw.sensors.8=viaenv0, VSENS3, volts_dc, 5.47 V
            hw.sensors.9=viaenv

      • The hardware sensors stuff is in there. Enable the options (i2c /dev interface if you want /dev nodes, or I think it might have been renamed in new kernels), install lm-sensors for the userspace stuff, and you can get things like gkrellm monitors if you want.

    • Re:About time (Score:5, Insightful)

      by Anonymous Coward on Monday March 27 2006, @06:55AM (#15001747)
      "Now Linux had this functionality how many years ago?"


      If by "functionality" you mean hodge-podge of barely compatible tools written by some high scool kid in his mum's basement and that fail to actually define a sensible engineered framework, then yes I suppose so. Jesus Tap Dancing Christ, Linux sucks ass.

      • Being all-powerful, I'm sure Jesus would be pretty good at tap dancing if he felt like it :-)

        That said, although 'lm_sensors' and such can be a royal pain to manage at a low level when starting out, many higher-level tools exist to manage entire networks of Linux machines and their status data.

        See the related apps page on the rrdtool [rrdtool.org] homepage.

    • Re:What about ACPI? (Score:4, Informative)

      by ThePhilips (752041) on Monday March 27 2006, @07:26AM (#15001833) Homepage Journal
      IIRC Intel's ACPI code was included in Kernel long time ago. It's just ACPI has nothing to do with sensors. (http://acpi.sourceforge.net/ [sourceforge.net])

      Sensors it's LM78 project. But. Not on single Linux instalation I've had luck with sensor installation. )-: Most of the time lm78 reported me nothing - given it found any sensors at all...

      P.S. Overall, due to separate development of kernel and libc, Linux development rarely results in any kind of API or framework. (Well, except the even rarer case when both developers - libc & kernel ones - happen to be employed by Red Hat.)
      • I don't remember just how I got this working in the first place, but lm_sensors works great on my Linux machine. I think it's supposed to work with some particular sensor models that don't go through ACPI... so if you've been doing all your Linux installations on machines with different types of sensors you'll probably have to look somewhere else. On the other hand, I have a laptop running FreeBSD and it seems like on there all the sensor-type stuff is handled through ACPI. So maybe if you have the right
    • Seriously. If my job wasn't so boring, I wouldn't give /. any of my time. But for now it serves an important purpose keeping me from going insane due to the tedium.

      I think I may code an AI script that will learn how to have conversations based on the content of slashdot. After the program has digested a few thousand posts it will surely pass the /. Turing test (Can a human distinguish this program from a typical /. poster?)

      I imagine a conversation would run like this:

      Human: "I'm impressed with this new Linu
    • ...quoth the anonymous coward. The irony is delicious, especially with steak sauce.

All trademarks and copyrights on this page are owned by their respective owners. Comments are owned by the Poster. The Rest © 1997-2009 Geeknet, Inc.