Skip to content

OS Level Virtualization Monitoring

OS Supported
FreeBSD jails
Linux cgroups v2(Docker, Podman included)

Install prereqs

apt-get install libjson-perl libclone-perl libmime-base64-perl libfile-slurp-perl libio-interface-perl cpanminus
cpanm OSLV::Monitor

pkg install p5-OSLV-Monitor

cpanm JSON Clone Mime::Base64 File::Slurp IO::Interface
cpanm OSLV::Monitor

SNMP Extend

  1. Setup cron.

    */5 * * * * /usr/local/bin/oslv_monitor -q > /dev/null 2> /dev/null

  2. Setup snmpd.

    extend oslv_monitor /bin/cat /var/cache/oslv_monitor/snmp

  3. Restart snmpd.

    sudo systemctl restart snmpd

Wait for it to be rediscovered by LibreNMS.

An optional config file may be specified via -f or placed at /usr/local/etc/oslv_monitor.json.

The following keys are used in the JSON config file.

Option Description Default
include An array of regular expressions to include. ["^.*$"]
exclude An array of regular expressions to exclude. undef
backend Override the backend and automatically choose it.
time_divider Override the time_divider value. see chapter under Time divider.

Time divider

The default value varies per backend and if it is needed.

cgroups

While the default for usec to sec conversion should be 1000000, some settings report the value in nanoseconds, requiring 1000000000.

Backend Time Divider Default
cgroups usec to sec 1000000
FreeBSD Not used

Backend

By Defaults the backends are as below.

Backend Default
FreeBSD FreeBSD
Linux cgroups

Default would be like this.

{
  "include": ["^.*$"]
}

Metric Notes

Key Description
running_$name 0 or 1 based on if it is running or not.
oslvm___$name___$stat The a specific stat for a specific OSLVMs.
totals_$stat A stat representing a total for all stats across all OSLVMs.

Something is considered not running if it has been seen. How long something is considred to have been seen is controlled by apps.oslv_monitor.seen_age, which is the number of seconds ago it would of have to be seen. The default is 604800 which is seven days in seconds.

All time values are in seconds.

All counter stats are per second values for that time period.

Backend Notes

FreeBSD

The stats names match those produced by ps --libxo json.

Linux cgroups v2

The cgroup to name mapping is done like below.

Input Output
systemd s_$name
user u_$name
docker d_$name
podman p_$name
anything else $name

The following ps to stats mapping are as below.

ps stats
%cpu percent-cpu
%mem percent-memory
rss rss
vsize virtual-size
trs text-size
drs data-size
size size

procs is a total number of procs in that cgroup.

The rest of the values are pulled from the following files with the names kept as is.

  • cpu.stat
  • io.stat
  • memory.stat

The following mappings are done though.

The following mappings are done though.

cgroupv2 stats
pgfault minor-faults
pgmajfault major-faults
usage_usec cpu-time
system_usec system-time
user_usec user-time
throttled_usecs throttled-time