System monitoring: A new frontend to log CPU, Memory and SWAP usage
Joseph McKenna
Branch: system_monitoring
system_monitoring
Branch: develop
develop
Merged
#8 · Created  · Last updated
Merged pull request
Merged in system_monitoring (pull request #8)
b96bda2·Author: Joseph McKenna·Closed by: Stefan Ritt·2019-10-18
Description
Â
Add prototype system monitor frontend that can log a systems CPU, Memory and SWAP usage. On start up it generates two user friendly history graphs based on the client hostname. The load from this frontend is extremely low
I have included it in the MIDAS repository as I think it should be an integral program that has an instance running on every DAQ machine in an experiment. No need to setup other systems such as ganglia to monitor and log system load. Feedback on this work is very welcome
This is only suitable for UNIX as it functions in the same ways as htop, parsing the /proc/ filesystem (specifically /proc/stat and /proc/memstat )
Â
Name: sysmon.cxx
Created by: J.T.K.McKenna
Contents: Front end for monitoring CPU and Memory usage with MIDAS
*
* Parse /proc/stat and /proc/memstat like htop
*
* I trust users to assign a new instance number for each machine they
* want to monitor... eg:
* ssh mydaq sysmon -i 1
* ssh myvme sysmon -i 2
* ssh mypi sysmon -i 3
Â
This frontend designed and tested to work with the current version of MIDAS (the front end doesnt have the C bindings of older code etc). Only small modifications are needed to bring it back
My own personal test setup:
I have tested the frontend on x86_64 (CENTOS 7 and Fedora 30), and ARMv7 raspberry pi systems (CENTOS 7), both logging to the ALPHAg DAQ:
Raspberry Pi logging example (I made the pi compile something to make the graph more interesting):
Â
Â
The frontend is limited to support upto 1000 CPUs in one computer
Â
Add prototype system monitor frontend that can log a systems CPU, Memory and SWAP usage. On start up it generates two user friendly history graphs based on the client hostname. The load from this frontend is extremely low
I have included it in the MIDAS repository as I think it should be an integral program that has an instance running on every DAQ machine in an experiment. No need to setup other systems such as ganglia to monitor and log system load. Feedback on this work is very welcome
This is only suitable for UNIX as it functions in the same ways as htop, parsing the /proc/ filesystem (specifically /proc/stat and /proc/memstat )
Â
Name: sysmon.cxx Created by: J.T.K.McKenna Contents: Front end for monitoring CPU and Memory usage with MIDAS * * Parse /proc/stat and /proc/memstat like htop * * I trust users to assign a new instance number for each machine they * want to monitor... eg: * ssh mydaq sysmon -i 1 * ssh myvme sysmon -i 2 * ssh mypi sysmon -i 3Â
This frontend designed and tested to work with the current version of MIDAS (the front end doesnt have the C bindings of older code etc). Only small modifications are needed to bring it back
My own personal test setup:
I have tested the frontend on x86_64 (CENTOS 7 and Fedora 30), and ARMv7 raspberry pi systems (CENTOS 7), both logging to the ALPHAg DAQ:
Raspberry Pi logging example (I made the pi compile something to make the graph more interesting):
Â
Â
The frontend is limited to support upto 1000 CPUs in one computer