Recent Changes - Search:
Site /

DownloadTheNjmonAndRelatedSourceCodeAndPre-compiledBinaryFiles

njmon for Linux with NVIDIA GPU support
  • 31 October 2019

Download:

  • njmon for Linux version 33 C code & Makefile njmon_linux_code_v33.zip - 26 KB single C program file and please ignore the Makefile (oops) it is below
  • njmon for Linux version 33 pre-compiled njmon_linux_rhel77_gpu.zip - 83 KB
  • Binary: Power ppc64le: RHEL 7.7 with GPUs install and the library

Makefile for the GPU version 

# Makefile for njmon for Linux
CFLAGS=-g -O4
LDFLAGS=-g

FILE=njmon_linux_v33.c
OUT=njmon

$(OUT):
        cc $(CFLAGS) -o $(OUT) $(FILE) $(LDFLAGS)

# Use the below to compile in the GPU support
# The libnvidia-ml.so or libnvidia-ml.so.1 may be in a different directory
# You will need to find it and replace the location below
# Then use:  make -f Makefile njmon_gpu
# Note: the resulting njmon_gpu will NOT run on a system without the library and supporting hardware
$(OUT)_gpu:
        cc $(CFLAGS) -o $(OUT)_gpu $(FILE) $(LDFLAGS) -D NVIDIA_GPU /usr/lib64/libnvidia-ml.so.1

clean:
        rm -fr $(OUT) $(OUT)_gpu
  • New features
    • Added Power (ppc64le) support for the NVIDIA P100 GPUs available on the AC922 under RHEL 7.7
      • To build and run in the GPU support you need the libnvidia-ml.so library installed. See the Makefile for details.
      • Once built with GPU support the njmon program will not on systems without the library
      • Stats include GPU description,GPU MHz, GPU Utilisation, GPU Memory Utilisation, GPU Temperature and GPU Watts (electrical power)
    • If you compile without the -D NVIDIA_GPU, you only get all the regular njmon functions.
  • Fixes
    • Improved -h help info.
  • IF you don't have NVIDIA GPUs - do not bother with this new version - the rest of the code is unchanged.
  • I suspect this would work on NVIDIA GPUs on AMD64/x86_64 but there is not way for me to tell - I assume there is a very similar library.
njmon for Linux
  • 28 October 2019

Download:

  • njmon for Linux version 32 C code & Makefile njmon_linux_code_v32.zip - 26 KB single C program file and Makefile
  • njmon for Linux version 32 pre-compiled njmon_linux_binaries_v32.zip - 600 KB
  • Binaries: x86_64: Ubuntu 18 & RHEL 7.5 and Power ppc64le: Ubuntu 18, RHEL 7.5, RHEL 8, SLES 12 and 15.
  • New features
    • NFS stats that are auto discovered and switched on
      • NFS v2 18 stats from client + 18 stats for servers
      • NFS v3 22 stats from client + 22 stats for servers
      • NFS v4 48 stats from client + 59 stats for servers
      • Note only really tested v4. Let me know if you use and it works for v2 and v3.
    • Converted hard stop errors in to warnings and try to carry on. Use a new option -w to stop warnings in the stderr channel.
    • Use "export NJMON_SECRET=password shell variable so that the password does not appear in the ps command output.
  • Fixes
    • Corrected a problem with -k mode. So it does actually stop quietly when the previous njmon is still running. This allows hour start attempts to reduce missing stats and reboot or filesystems filling.
    • Deprecated -M (default) and -S options - let me know if you use -S (single level stats output).
njmon for AIX 16 Oct 2019

Download:

  • njmon for AIX/VIOS version 34 C code & Makefile njmon_aix_code_v34.zip - 31 KB single C program file and Makefile
  • njmon for AIX/VIOS version 34 pre-compiled njmon_aix_binaries_v34.zip - 1.2 MB
  • Fixes
    • Fixes the -k option (is was saving the wrong pid = the parent).
  • Additions
    • This release supports NFS versions 2, 3 and 4
    • Numbers of Stats: NFS v2 = 18 (both client & server), NFS stats v3 = 22 (both client & server) and NFS v4 server side 49 and client side 41
    • If AIX does to supply any NFS stats of a particular version or if the stats are all zeros then nothing is output - to reduce capturing pointless stats.
njmon for AIX 16 Sept 2019

Download:

  • njmon for AIX/VIOS version 33 C code & Makefile njmon_aix_code_v33.zip - 30 KB single C program file and Makefile
  • njmon for AIX/VIOS version 33 pre-compiled njmon_aix_binaries_v33.zip - 1.2 MB
  • Fixes
    • Big bug corrected in a memory copy function (memcpy() and pointer), thanks to Melaku Tena Wondimu who works at the Commercial Bank of Ethiopia for the feedback and fix (removing two "&" characters but gets the correct memory copied)
      • This might explain a few flakey issues in the past in tape, paging space and netbuf stats.
      • Might be worth revisiting these stats if you switched them off.
    • Disk Adapter types should be reported correctly now as "SCSI, SAS, other" or "Virtual SCSI/SAS Adapter" or "Fibre Channel" (no longer all as "unknown")
  • Changes:
    • CPU number for Logical CPU and Physical CPU are now cpu001, cpu002 etc so they can be lists in numeric order on graphs. Previous we get cpu1, cpu11, cpu12, cpu2 etc.
    • Added partition type, hopefully, it is obvious what these mean
      • Except "shared_extended" as the documentation appears missing:
             "partition_type": {
          "smt_capable": 1,
          "smt_enabled": 1,
          "lpar_capable": 1,
          "lpar_enabled": 1,
          "shared_capable": 1,
          "shared_enabled": 1,
          "dlpar_capable": 1,
          "capped": 0,
          "kernel64bit": 1,
          "pool_util_authority": 0,
          "donate_capable": 1,
          "donate_enabled": 0,
          "ams_capable": 1,
          "ams_enabled": 0,
          "power_save": 1,
          "shared_extended": 0
       }
    • Added partition_pool stats for the NEW Shared Processor Pool stats that the LPAR running njmon is in.
      • Note you need the HMC LPAR setting: Enable Performance Information Collection or the HMC CLI lssysvfg LPAR field=allow_perf_collection = 1
      • njmon should cope with you switching this option on, while it is running (no need to restart njmon).
      • I think the most interesting is the 'busy_cores of the pool.
      • This data should be very useful of software licence tracking where software is priced/licensed per CPU and a CPU pool used prove use. njmon is not an IBM legally approved tool.
      • It is obvious Shared Processor Pools don't work for Dedicated CPU LPARs, right! 
             "processor_pool": {
          "pool_id": 1,
          "max_capacity": 10,
          "entitled_capacity": 7,
          "phys_cpus_pool": 10,
          "idle_cores": 8.683,
          "max_cores": 9.996,
          "busy_cores": 1.305,
          "scaled_busy_cores": 1.300,
          "global_pool_tot_cores": 31.987,
          "global_pool_busy_cores": 3.987
     }
    • Added further runtime ASSERT checks
    • Withdrawing:
      • Not going to support the older JSON formats of single-level (option -S) JSON as these don't work for TimeSeries insertion like InfluxDB
      • This time removed from the njmon -h info but still work.
      • Next time removed to a compile-time option
      • Reason: I don't think anyone uses this format - shout if you want them!
Phase 3 = a Major Update - All version numbers are moved to 30 to make this clear
  • Due to being new to InfluxDB/Grafana I made a mistake that phase 3 corrects.
  • Better use of tags means we can have graphs in a template that is not fixed for a particular server/VM
  • For resources that have subresources like CPUs, Disks, networks, files systems we need to put the CPU numbers, disk names, network names, etc. into the tags when injecting into InfluxDB
  • Then a small change to Grafana graphs definitions does all the work for us.
  • See below using filesystems as an example:
  • We can also do iceberg graphs read above the zeroth horizontal line and write below zero makes it easier to see how they change:
  • Below we do maths to add User+System+wait Utilisation with a GroupBy by hitting the Edit Mode icon
njmon for AIX 17 July 2019

Download:

  • njmon for AIX/VIOS version 31 C code & Makefile njmon_aix_code_v31.zip - 29 KB single C program file and Makefile
  • njmon for AIX/VIOS version 31 pre-compiled njmon_aix_binaries_v31.zip - 1.2 MB
  • Fixes
    • Really annoying mistake that had OLDNAMES compiler directive was switch on - fortunately, the Injector corrects this so no damage.
    • Correct a "network_interface" bug - the 1st stats could be in the quadrillions - the network I/O since boot time! This mucks up graphs as it rescales for the massive numbers.
    • Correct the "cpu_logical" calculation that now avoids an integer maths transitory result is a faction (and so a zero integer).
      • Note:t if all the CPU logical numbers are zero it probably means the logical CPU is off (not being scheduled).
    • Logical and Physical CPU names are now "cpuN" as returned from the libperfstat library and not "lcpu" and "pcpu" - might get interesting if you dynamically change the numbers of CPUs.
    • Rename "identity" command and version to njmon_command and njmon_version to be clearer and the same as njmon for Linux
  • 10 July 2019
  • Withdrawn: njmon for AIX/VIOS version 30 C code & Makefile
  • Withdrawn: njmon for AIX/VIOS version 30 pre-compiled
  • Changes:
    • JSON files have the pointed "sample": { . . . } removed - it was pointless.
      • So the JSON file contains an array of JSON structures { [{ }, {}, . . . {} ] }
      • This does not affect the Injector but other tools might be affected
    • AIX data typo _agv changed to _avg - short for average
    • AIX rename some JSON section titles so the all the CPU stats appear together in an alphabetically sorted list in Grafana
      1. logical_cpu to cpu_logical
      2. physical_cpu to cpu_physical
      3. total_logical_cpu to cpu_logical_total
      4. total_physical_cpu to cpu_physical_total
      5. total_physical_cpu_spurr to cpu_physical_total_spurr
      6. syscals to syscalls
    • Big Change due to my attending an InfluxDB conference and realising we are not using tags properly
      • For subsections like "disks": ("hdisk0":{}, "hdisk1": {} etc the tags should point of the disk names then we can use group-by operations in Grafana.
      • Then we have "tags": {"disk_name": "hdisk0" }
      • The result is the Grafan templates can graph any number of sub-devices automatically.
      • The same goes for disk adapters, logical & physical CPUs, network adapters and interfaces, file systems, memory page sizes, paging spaces.
    • Now the bad news - it is far better to start again reloading saved JSON files with the new Injector with improved tag
    • NO LONGER NEEDED: So STOP using separate AIX data and Linux data databases
      • Just one database for both AIX and Linux
    • Below changed to just one database called njmon30
      • For example: 
        $ influx
Connected to http://localhost:8086 version 1.7.1
InfluxDB shell version: 1.7.6
Enter an InfluxQL query
> create database njmon30
> show databases
name: databases
name
----
_internal
njmon14    -> This is my older database
njmon30
nextract
> exit
$
    • NO LONGER USE TWO COLLECTORS
  • Simplified picture with one collector, one injector and one Influx database:
njmon collector 10 July 2019

Download:

  • Version 30 withdrawn - bug due to confusing the collector version 30 and the protocol version 12 between njmon and the collector
  • Please update to 31 ASAP to avoid collector 100% failures
  • njmon collector version 31 njmon_collector_v31.c - C source code
  • Compile with: cc -O4 -o njmon_collector_v31 njmon_collector_v31.c
  • njmon collector version 30 njmon_collector_v31_binaries.zip
    • Binary for Ubuntu 18.04 on AMD64 (x86_64) and RHEL7.6 on ppc64le (POWER9)
  • New Features
    • Option to use a config file instead of lots of command line parameters.
    • This means passwords are not showing up on "ps -ef" output.
    • So njmon_collector -a /home/nag/collector.conf
    • Content: 
      port=1234
directory=/home/nag/njmondata
secret=abcd1234
inject=1
injector=/usr/local/bin/njmon_influxDB_injector_online_mode.py
json=1
    • The others changes are minor.
njmon injector 10 July 2019
  • New Features
    • Using better tags for handling subsections
    • Improved logging
  • njmon influxDB injector njmon_influxDB_injector_30.py - Python program
  • DO NOT FORGET TO CHANGE THE SETTING AT THE TOP OF THIS PROGRAM
  • The first to point at the InfluxDB server
  • Example in a file called inject_batch.py:
     
 host="localhost"
 port=8086
 user = 'Nigel'
 paassword = 'SECRET'
 dbname = 'njmon30'

 batch=True
  • Example in a file called inject_online.py:
     
 host="localhost"
 port=8086
 user = 'Nigel'
 paassword = 'SECRET'
 dbname = 'njmon30'

 batch=False
  • Set to batch=True (for loading a completed JSON file quickly) and batch=False (for live loading JSON records as they arrive).
njmon for Linux 16 July 2019

Download:

  • njmon for Linux version 30 C code & Makefile njmon_linuxcode_v30.zip - 23.4 KB single C program file and Makefile
  • njmon for Linux version 30 pre-compiled njmon_linux_binaries_v30.zip - 0.7 MB
  • njmon for Linux version 30 JSON sample files njmon_linux_JSON_samples_v30.zip - 0.1 MB
  • Fixes
    • Reworked the measurements of sleep time overrun (normally when CPU very busy) and execute time extending (more as the resources get larger like CPU, Disks, Networks, filesystems and if Process stats are collected).
  • Improvements:
    • Like njmon for AIX removed the: "sample": { and ending: }
      • These lines were pointless and added complexity reading the JSON data.
    • Statistics "stat" (not a helpful name) are from the proc/stat
      • This file contains three different sorts of stats that should not have been mixed up (sorry about that), if we want to graph them. So they are renamed:
      • "stat" + CPU total line becomes "cpu_total"
      • "stat" + one or more individual CPUs becomes "cpus" and cpuNNN (NNN is the CPU number starting from zero)
      • "stat" + ctxt (context switches), btime (boot time seconds since Epoch), processes, procs_running and procs_blocked becomes "stat_counters"
    • Section "disk" renamed "disks" to follow the convention like all the others with subsections
    • Changes indent level for subsection names for better readability. For example, "cpu0" : { is one Tab character to the left.
    • Important: See above for details of creating a NEW InfluxDB - influx commands create database. I suggest njmon30linux for a database name.
    • Changes in njmon for Linux and in the InfluxDB injector means the data is incompatible - so a fresh start is needed, sorry.
njmonchart 17 July 2019

Download:

  • njmonchart for AIX and Linux version 30 Python3 programs njmonchart_aix_linux_v30.zip - 14 KB
  • Changes for the new njmon for AIX v30 and njmon for Linux v30 like the removal of "sample": and improved section names.
  • Tried to maintain backward compatibility to handle older files but you may be better off using the older njmonchart for them.
  • njmonchart for Linux has new graphs - CPU Busy and MHz for each CPU
Older Versions
njmon for AIX 6 June 2019

Download:

  • njmon for AIX/VIOS version 22 C code & Makefile njmon_aix_code_v22.zip - 29 KB single C program file and Makefile
  • njmon for AIX/VIOS version 22 pre-compiled njmon_aix_binaries_v22.zip - 1.2 MB
  • Odd startup bug that affected about 30% AIX/VIOS LPARs every time but not others.
    • It outputs an ASSERT "ERROR: Reason=Elapsed Calculation" and stops at the first loop i.e if the option "-s 60" is used it forces a core dump after 60 seconds.
    • Test this fix on 7 AIX and VIOS LPAR - looks good.
  • 22 May 2019
  • njmon for AIX/VIOS version 21 C code & Makefile njmon_aix_code_v21.zip - 29 KB single C program file and Makefile
  • njmon for AIX/VIOS version 21 pre-compiled njmon_aix_binaries_v21.zip - 1.2 MB
  • For
    • AIX 6.1, AIX 7.1, AIX 7.2 TL2, AIX 7.2 TL3
    • VIOS 2.2, VIOS 3.1
  • Fixed
  • New Features
    • Snapshot time creep - njmon can on busy VMs snapshot stats slightly late and so 60-second snapshots and 1440 captures take slightly longer than a day by a few minutes.
      • njmon monitors how long it sleeps between snapshots (on a very busy server a 30 sleep might take (say) 30.3 seconds as all CPUs are busy and there is a long run queue)
      • njmon monitors how long it takes to captures the stats (can add up on larger VM configurations and if you ask for extra stats like processes).
      • This "extra" time is accumulated and once it reaches 1 second the next (for example) 30 sleep is reduced to a 29-second sleep to keep the snapshots regular.
  • More useful comment in ASSER failures - like njmon with shutdown a stats but carry on.
  • 25th March 2019
  • njmon for AIX/VIOS version 20 pre-compiled njmon_aix_binaries_v20.zip - 1.2 MB for AIX 6.1, AIX 7.1, AIX 7.TL2, AIX 7.2 TL3 plus VIOS 2.2, VIOS 3.1
  • njmon for AIX/VIOS version 19 C code njmon_aix_code_v20.zip - 29 KB single C program file and Makefile
  • Fixed Disk Busy% stats.
  • Note: the below requires the -P option to switch on process stats
  • Note: process CPU times are in a percent of the LPARS whole CPU time - unlike nmon which is a percent of a single CPU core.
  • Added Asynchronous I/O stats = a summary of the aioserver processes (like nmon) see aioserver stats section with stats aioprocs, aiorunning and aiocpu. 
  •      "aioserver": {
          "aioprocs": 4,
          "aiorunning": 4,
          "aiocpu": 5.298
     }
  • 5.298% is out of the 100% for the LPAR.
  • Added CPU ignore threshold to not report processes that are not using CPU time. The default is 0.001% of the LPAR's CPU.
  • Added -t percent option to change the threshold.
  • Removed in the code STAY_FOREGROUND compiler directive. Use -D command-line option instead.
  • njmon for AIX Command line options and njmon -? output: njmon4AIXCommandLine
njmon for Linux 30 May 2019

Download:

  • njmon for Linux version 26 C source code and Makefile: njmon_linux_code_v26.zip - 23 KB.
  • njmon for Linux version 26 pre-compiled code - njmon_linux_binaries_v26.zip - 565 KB.
    • POWER ppc64le (Ubuntu18.04, RHEL 7.5, RHEL 8, SLES 15)
    • AMD64/x86_64 (Ubuntu18.04, RHEL 7.5)
    • ARM64 (Ubuntu16 on Raspberry Pi)
  • Features
    • Bug fix: remove the divide by zero on automount NFS directories reporting zero size
    • Bug fix: wrong index used in Top Process processing. This will only effect you, if you are collecting Top Process stats (-P option).
    • A big thanks to Mr. Bruce Howells for finding these and even suggesting fixes.
  • 22 May 2019
  • njmon for Linux version 25 C source code and Makefile: njmon_linux_code_v25.zip - 23 KB.
  • njmon for Linux version 25 pre-compiled code njmon_linux_Binaries_v25.zip - 500 KB.
    • POWER ppc64le (Ubuntu18.04, RHEL 7.5, SLES 15)
    • AMD64/x86_64 (Ubuntu18.04, RHEL 7.5)
    • ARM64 (Ubuntu18 on Raspberry Pi)
  • New Features
    • Snapshot time creep - njmon can on busy VMs snapshot stats slightly late and so 60-second snapshots and 1440 captures take slightly longer than a day by a few minutes.
      • njmon monitors how long it sleeps between snapshots (on a very busy server a 30 sleep might take (say) 30.3 seconds as all CPUs are busy and there is a long run queue)
      • njmon monitors how long it takes to captures the stats (can add up on larger VM configurations and if you ask for extra stats like processes).
      • This "extra" time is accumulated and once it reaches 1 second the next (for example) 30 sleep is reduced to a 29-second sleep to keep the snapshots regular.
    • Network stats - both RHEL and SLES newer releases don't default install ifconfig (crazy IMHO)
      • Coded around this "mistake" and actually improved the code so later added (i.e. while njmon is running) network adapters are captured.
  • OS version name capture
    • Further updates to work around Linux nonsense - Some Linux's have trailing spaces in the /etc/os_release file.
    • The bug cause njmon to record things like "OS_name": "Oracle Linux version"" (double double-quotes)
  • 16 April 2019
  • Note: withdrew v23 to remove a typo and it was reported Oracle Linux has some stray spaces at the end of the lines in /etc/os-release - fixed in v24.
  • njmon for Linux version 24 C source codenjmon_linux_v24.c - 58 KB.
  • njmon for Linux version 24 pre-compiled To Be Released Soon for
    ppc64le (Ubuntu18.04, RHEL 7.5) and AMD64/x86_64 (Ubuntu18.04, RHEL 7.5)
  • New Features:
  • -P added Top Process stats: includes memory stats (if run as root), includes user id and user name.
    • Warning: this on busy systems will take extra CPU time
  • -I lets you change the default ignore threshold of 0.01 % of a CPU (not normally needed).
  • Change JSON indents to Tabs and not 4 spaced for reduced output size but keep user-friendly formatting.
  • Fixes to get Red Hat RHEL 5 and RHEL 6 working with the recent injector that demands OS version stats for tags.
  • Sample process stats: 
 
"processes": {
    "process_6649": {
        "pid": 6649,                   "cmd": "influxd",                   "ppid": 1,                        "pgrp": 6649,
        "priority": 20,                "nice": 0,                          "session": 6649,                  "tty_nr": 0,
        "flags": "0x40400100",         "state": "Sleeping-interruptible",  "threads": 34,                    "cpu_percent": 0.899,
        "cpu_usr": 0.883,              "cpu_sys": 0.017,                   "cpu_usr_total_secs": 56035.010,  "cpu_sys_total_secs": 3735.470,
        "statm_size_kb": 10568644,     "statm_resident_kb": 435040,        "statm_restext_kb": 11436,        "statm_resdata_kb": 5714716,
        "statm_share_kb": 25316,       "minorfault": 927828.505,           "majorfault": 19.988,             "it_real_value": 0,
        "starttime_secs": 2340018.860, "virtual_size_kb": 10568644,        "rss_pages": 108760,              "rss_limit": -1,
        "swap_pages": 0,               "child_swap_pages": 0,              "last_cpu": 2,                    "realtime_priority": 0,
        "sched_policy": 0,             "delayacct_blkio_secs": 0.000,      "uid": 999,                       "username": "influxdb"
        },
        ....
}
njmonchart for Linux 21st May 2019
  • njmonchart for Linux version 9 here: njmonchart_linux_v9.py 25 KB
    • Added Logical CPU thread (Logical CPUs) graphs
      • Historgram showing each Logical CPU use - This for PowerVM highlights SMT use (or not)
    • Added Specturm Scale (GPFS) for graphs:
      • Whole GPFS Read & Write bytes/s
      • Whole GPFS Each Operation/s
      • Individual filesystems Read & Write bytes/s
      • Individual filesystems Each Operation/s
    • Fixes
      • Handing serial-number and serial_number in JSON files (this was changed earlier)
      • Correction Memory stats Units
  • njmonchart for Linux version 7 here: njmonchart_linux_v7.py 19 KB
  • 28th April 2019
  • With a starter pack set of nmonchart graphs, including:
    • Configuration,
    • Total CPU, Run Queue, pSwitch, Forks,
    • Memory, PageSpace, Paging,
    • Total Disk RW, Total Disk Xters, Disk-Busy, Disk-KB, Disk-Ops,
    • JFS,
    • Total Net-Bytes, Total Net-Xfer, Net-MB, Net-packets
  • All in 480 lines of Python code.
  • Consider this v1 starter pack based on the njmonchart for AIX v7 code
  • Further test is recommended and suggestions of the most important graphs would be useful to focus on improvements.
njmonchart for AIX 21st May 2019
  • njmonchart for AIX/VIOS version 9 here: njmonchart_aix_v9.py 33 KB
    • Added Logical CPU thread (Logical CPUs) graphs
      • Historgram showing each Logical CPU use - This for PowerVM highlights SMT use (or not)
    • Added Specturm Scale (GPFS) for graphs:
      • Whole GPFS Read & Write bytes/s
      • Whole GPFS Each Operation/s
      • Individual filesystems Read & Write bytes/s
      • Individual filesystems Each Operation/s
  • njmonchart for AIX/VIOS version 7, here: njmonchart_aix_v7.py 27 KB
  • 11th Feb 2019
    • With a full set of nmonchart graphs, including Configuration, Top Process bubble chart and top 20), Top 20 disks,
      CPU (many), Disks (many), networks (many) IPC, plus a few new ones: CPU Mhz, Overclocking %, Load Avg numbers,
      Interrupts, Decrementor Interrupts, Memory Zero and Namei (directory/ inode lookups).
    • All in 600 lines of Python code.
    • Small Sample on njmonchart v7 output. It is a web page with graphs: njmonchart_silver5_sample.html 72 KB
njmon Injector for InfluxDB 11 March 2019
  • njmon Injector for InfluxDB version 15 for AIX and Linux
    • This version adds architecture, serial number and Machine-Type_Model for Linux data tags to aid Grafana Templates
    • njmon_to_InfluxDB_injector_15.py new injector.py for njmon JSON files into InfluxDB.
    • You MUST edit injector file for
      • the InfluxDB hostname
      • the InfluxDB to use
      • The Username and password
    • Supplied for batch bulk injection.
    • NOTE: If you want to single snapshot injection for live "as you collect the data" mode then change batch=True to batch=False
    • Added logging for debugging - it will create an injector.log file to the directory in which the njmon_collector is running and
      add one log line each time it pushes data into InfluxDB
njmon Collector 15 March 2019
  • Version 16
  • Single new feature - If injecting you can switch off saving the JSON file using the -n option- so we don't fill up the disk
Grafana Template
for njmon for AIX 		1 Feb 2019
njmon splunk injector Nov 2019
  • Injectors for adding njmon JSON data live into InfluxDB (and Grafana) or Splunk - these also work for njmon files in JSON format

Older Versions - Only use these if Nigel makes a ghastly mistake in a newer version

Grafana Graph Templates

  • The grafana.com has a mechanism to uploading and downloading template files
  • These are JSON files you can import
  • There are currently 4 for graph templates for njmon - you can search for them
  • One above
    • Grafana_Template_for_njmon_AIX_v3-1548086037850.json
  • Sascha Wysick

- - - The End - - -