Bugfix release

• Fix regression introduced in v1.7-0: when collecting per-process stats, only the main thread of each process was considered.
• Many improvements to charts generated for cgroups/containers
• Improvements to the versioning of the build
• Add "cpu_tot" measurement under "cgroup_cpuacct_stats" section to ease post-processing scripts
• Add CPU throttle and memory failure allocations reporting to cmonitor_statistics and make it more reliable
• Fix detection of "no-limits" for CPU/memory in cgroups

Feature release

This release contains a new powerful feature: per-thread CPU/IO resource collection:

• add new 'cgroup_threads' --collect option value to trigger collection of stats associated with secondary threads of processes found inside the cgroup-under-monitor
• add new '--score-threshold' option value to make configurable the score filter for monitored processes/threads; this is useful to force cmonitor_collector to collect and save stats for all threads, regardless of their CPU usage.

Feature release

Feature release

• add --custom-medata option to allow saving whatever information associated with the sampled data into the JSON and use it in cmonitor_chart
• add --utc option to cmonitor_chart to plot data in UTC time instead of local timezone
• add --remote-dbname option to avoid hardcoding 'cmonitor' influxDB database name (#8)
• add cmonitor statistic tool to generate statistics summary including mean, median and mode
• add recording of CPU throttling in cmonitor_collector
• split the RPM in 2 RPMs: cmonitor_collector and cmonitor_tools; in this way the first RPM has no Python3 dependency
• add recording of "CPU quota limit" in cmonitor_collector
• add plot of CPU limit, memory limit and CPU throttling

Bugfix release

• Bugfix in cmonitor_chart for the case only cgroup statistics (no baremetal stats) are collected

Small bugfixes for cgroup CPU accounting

• Fix regression in CPU accounting slipped in 1.4-1

Small improvements and bugfixes

Improve Python script CLI interface (provide --help, --verbose, --input and --output) in cmonitor_chart.
Add --num-samples=until-cgroup-alive to cmonitor_collector.
Bugfixes (notably: fix cpuacct collection)

Add support for monitoring any container from the baremetal

• Add --cgroup-name option to allow monitoring cgroups that are not necessarily the one where cmonitor_collector runs.
• Switch from Centos7 as official Docker base to Alpine to get a much smaller docker image.
• Refactor example section of the README to better show the possible usage

Add support for process statistics

This release introduces per-process statistic collection and charting.
The CPU/memory/disk performance of all processes that belong to the sampled cgroups are now monitored. The collected data can then be plotted by using cmonitor_chart Python tool. See https://f18m.github.io/cmonitor/examples/baremetal1.html for an example.

Add support for compressed JSON data in HTML

cmonitor_collector.py:
Add use of Pako JS zlib implementation to inflate datatables, now
compressed at JSON2HTML conversion time to save disk space and
bandwidth; new compressed baremetal2.html is 16% the size of the
original!
Add support for gzipped JSON stats files.

First version

First release after project rename (from "nmon-cgroup-aware" to "cmonitor").
All distributions (RPM, Ubuntu PPA, Docker) have been re-tested successfully.
Website updated.