Details

    • Improvement
    • Resolution: Fixed
    • Minor
    • None
    • None
    • OS, Unix Agent
    • None

    Description

      Hi, obeservium team!

      Many thanks for your great project.

       

      Could you please add following features to observium:

      • visibility of GPUs in linux host inventory
      • monitoring/graphing GPU load, we're interested in parameters "nvenc" and "nvdec" from "nvidia-smi dmon" command

      Attachments

        Activity

          [OBS-3588] Nvidia GPU + Linux

          Would be nice, if you will add enc and dec values.. They are very important in video processing/encoding business.

          ugenk Evgeniy Kozhuhovskiy added a comment - Would be nice, if you will add enc and dec values.. They are very important in video processing/encoding business.

          Hi, no.

          Only GPU environment sensors:

          nvidia-smi --format=csv --query-gpu=index,gpu_name,temperature.gpu,fan.speed,power.draw,utilization.gpu,utilization.memory

          landy Mike Stupalov added a comment - Hi, no. Only GPU environment sensors: nvidia-smi --format=csv --query-gpu=index,gpu_name,temperature.gpu,fan.speed,power.draw,utilization.gpu,utilization.memory

          Hi!

           

          I've checked around. Did you've added support for enc(oder) and dec(oder) variables?

          ugenk Evgeniy Kozhuhovskiy added a comment - Hi!   I've checked around. Did you've added support for enc(oder) and dec(oder) variables?

          SHould be fixed in r12202.

          landy Mike Stupalov added a comment - SHould be fixed in r12202.

          Hi!

           

          encoder.stats.sessionCount

          encoder.stats.averageFps

          encoder.stats.averageLatency

           

          also, i didn't found following paramenters, but we also need enc% and dec% from here:

          root@trscd17:~# nvidia-smi dmon
          # gpu   pwr gtemp mtemp    sm   mem   enc   dec  mclk  pclk
          # Idx     W     C     C     %     %     %     %   MHz   MHz
              0    49    61     -    27    15    65    19  3802  1822
              1    44    49     -    11     6    20     9  3802  1835
              0    53    62     -    32    17    62    24  3802  1822

           

           

          ugenk Evgeniy Kozhuhovskiy added a comment - Hi!   encoder.stats.sessionCount encoder.stats.averageFps encoder.stats.averageLatency   also, i didn't found following paramenters, but we also need enc% and dec% from here: root@trscd17:~# nvidia-smi dmon # gpu pwr gtemp mtemp sm mem enc dec mclk pclk # Idx W C C % % % % MHz MHz 0 49 61 - 27 15 65 19 3802 1822 1 44 49 - 11 6 20 9 3802 1835 0 53 62 - 32 17 62 24 3802 1822    

          Hi Evgeniy,

          I assume you're using the nvidia-smi agent script on these machines. I wrote that a long time ago, I'm not sure I have any Linux machine with Nvidia graphics lying around - I could take a look.

          If you run this command, does it contain the information you need? It's quite likely Observium is getting the info but just not doing anything with it yet (except for temperatures).

          nvidia-smi --format=csv --query-gpu=index,gpu_name,temperature.gpu,fan.speed,power.draw,utilization.gpu,utilization.memory
          

          Thanks!

          sid3windr Tom Laermans added a comment - Hi Evgeniy, I assume you're using the nvidia-smi agent script on these machines. I wrote that a long time ago, I'm not sure I have any Linux machine with Nvidia graphics lying around - I could take a look. If you run this command, does it contain the information you need? It's quite likely Observium is getting the info but just not doing anything with it yet (except for temperatures). nvidia-smi --format=csv --query-gpu=index,gpu_name,temperature.gpu,fan.speed,power.draw,utilization.gpu,utilization.memory Thanks!

          For now, only temperature is shown for GPU

           

          ugenk Evgeniy Kozhuhovskiy added a comment - For now, only temperature is shown for GPU  

          snmpwalk from linux host with two GPUs attached

          [^myagent.snmpwalk]

          ugenk Evgeniy Kozhuhovskiy added a comment - snmpwalk from linux host with two GPUs attached [^myagent.snmpwalk]

          Please make and attach additional information about the device:

          • full snmp dump from device:

            snmpwalk -v2c -c <community>  --hexOutputLength=0 -ObentxU <hostname> .1 > myagent.snmpwalk
            snmpwalk -v2c -c <community>  --hexOutputLength=0 -ObentxU <hostname> .1.3.6.1.4.1 >> myagent.snmpwalk

            If device not support SNMP version 2c, replace -v2c with -v1.

          • If you have problems with discovery or poller processes, please do and attach these debugs:

            ./discovery.php -d -h <device>
            ./poller.php -d -h <device>

          • additionally attach device and/or vendor specific MIB files

          Note, this comment is added automatically.

          bot Observium Bot added a comment - Please make and attach additional information about the device: full snmp dump from device: snmpwalk -v2c -c <community> --hexOutputLength=0 -ObentxU <hostname> .1 > myagent.snmpwalk snmpwalk -v2c -c <community> --hexOutputLength=0 -ObentxU <hostname> .1.3.6.1.4.1 >> myagent.snmpwalk If device not support SNMP version 2c, replace -v2c with -v1. If you have problems with discovery or poller processes, please do and attach these debugs: ./discovery.php -d -h <device> ./poller.php -d -h <device> additionally attach device and/or vendor specific MIB files Note, this comment is added automatically.

          People

            sid3windr Tom Laermans
            ugenk Evgeniy Kozhuhovskiy
            Votes:
            2 Vote for this issue
            Watchers:
            7 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved: