Docker-py and os.cpu_count() disagree on the count, leading to data
going beyond the scale of the graph. As Docker-py seems to have dummy
CPU data in the stats, we instead always rely on Python's CPU count.
Signed-off-by: Olivier Mehani <shtrom@ssji.net>
* Add support for docker_network
* Actually show non running containers counts
* Show intermediate and dangling images counts
* Add warnings on dangling images and dead containers
* Fully support autoconf and suggest
* Add graph_total to the status (thereby making the containers graph
redundant), CPU and Memory graphs
* Show most graphs as AREASTACK, this allows to get a clearer view of
both individual and total use
* Use LINESTACK1 to work around munin-contrib/munin#1343
* Include all containers in CPU/Memory/Network config, so data from
non-running containers is still displayed
* Sprinkle some info and extinfo
* Add ClientWrapper around docker module, providing caching and sorting
* Reverse-engineer author list from git log
It maybe desirable to not generate stats for ephemeral containers.
This commit adds an env variable EXCLUDE_CONTAINER_NAME which is a regex, if it matches
then the container is excluded from the cpu and memory graphs.
Added parallel processing using python's muliprocessing module
to fork a new process each for each docker container when gathering
stats.
This vastly improves performance since using a for loop the script
blocks until the stats are returned.
