newrelic / check_docker Goto Github PK

View Code? Open in Web Editor NEW

127.0 127.0 31.0 37 KB

A Go Nagios check for Docker

License: BSD 2-Clause "Simplified" License

Go 95.29% Shell 4.71%

check_docker's People

Contributors

Stargazers

Watchers

check_docker's Issues

No ability to check separate containers with same image name

Given a docker ps like this:

CONTAINER ID        IMAGE                                                        COMMAND             CREATED             STATUS              PORTS                  NAMES
edefe9ea93f8        foo/baz:59977daa1075abda15d7495781c1cf458e0cd863   script/run          3 days ago          Up 3 days           0.0.0.0:8485->80/tcp   dreamy_archimedes
3e656b925a5c        foo/baz:59977daa1075abda15d7495781c1cf458e0cd863   script/run          3 days ago          Up 3 days           0.0.0.0:80->80/tcp     romantic_albattani

There's no way to write a check that validates both of these containers are up. Passing -image-id="foo/baz" returns OK: 2 containers up, but would return OK even if only one was.

May not work on AUFS backends

Some of the stats being checked may only apply on devicemapper backends. Needs verification.

Connect to docker over unix domain socket

It would be nice if the check_docker was able to connect to the docker service over the unix domain socket; /var/run/docker.sock.

By default docker does not listen to a port, but uses the more secure connection through unix domain socket.

Sometime docker does not respond (unix socket in my scenario). When this happens, check_docker hangs forever. As we would like it to be a critical, for now we have to use a small wrapper to detect the timeout and raise the CRITICAL :

#!/bin/sh

timeout 5 /opt/check_docker/check_docker --base-url='unix:///var/run/docker.sock' --crit-data-space=90 --warn-data-space=85 --crit-meta-space=90 --warn-meta-space=85
R=$?
if [ "$R" = '124' ]; then
  echo "CRITICAL docker timeout"
  exit 2
fi

exit $R

I think it is a must-have for a check. We are using sensu here, but timeout is also a guideline for nagios checks https://nagios-plugins.org/doc/guidelines.html#RUNTIME

Check returns incorrect Metaspace utilisation due to human readable units

When checking against our docker systems check_docker is returning the following:
CRITICAL: 17 containers - Meta Space Used: 818% - Meta Space Used: 818%
Docker info reports the following:

    Data Space Used: 6.124 GB
    Data Space Total: 107.4 GB
    Metadata Space Used: 17.56 MB
    Metadata Space Total: 2.147 GB

this is likely due to the docker devmapper/driver.go converting to human friendly units as per below:

        status := [][2]string{
        {"Pool Name", s.PoolName},
        {"Pool Blocksize", fmt.Sprintf("%s", units.HumanSize(float64(s.SectorSize)))},
        {"Backing Filesystem", backingFs},
        {"Data file", s.DataFile},
        {"Metadata file", s.MetadataFile},
        {"Data Space Used", fmt.Sprintf("%s", units.HumanSize(float64(s.Data.Used)))},
        {"Data Space Total", fmt.Sprintf("%s", units.HumanSize(float64(s.Data.Total)))},
        {"Data Space Available", fmt.Sprintf("%s", units.HumanSize(float64(s.Data.Available)))},
        {"Metadata Space Used", fmt.Sprintf("%s", units.HumanSize(float64(s.Metadata.Used)))},
        {"Metadata Space Total", fmt.Sprintf("%s", units.HumanSize(float64(s.Metadata.Total)))},
        {"Metadata Space Available", fmt.Sprintf("%s", units.HumanSize(float64(s.Metadata.Available)))},
        {"Udev Sync Supported", fmt.Sprintf("%v", s.UdevSyncSupported)},

Given that human readability is hard coded in the driver, it seems that the plugin needs to take this into account and normalise all units for comparison.

Doesn't build with go 1.3 or 1.6 on Debian 8.3

I tried to build check_docker with go 1.3 standard Debian wheezy package installed and it failed. I downloaded and installed go 1.6 from golang.org and got the same result, as below.

$ go get github.com/newrelic/check_docker
# github.com/newrelic/check_docker
src/github.com/newrelic/check_docker/check_docker.go:68: cannot assign *docker.DockerInfo to cd.dockerInfoData (type *docker.Env) in multiple assignment

The line is part of the 5 Feb 2015 refactor and there are binary downloads after that, so clearly it (used to) build for others.

https support

Our docker host is secured by TLS as mentioned here. When using curl to access this secured docker daemon, these additional switches are required:

$ curl --insecure --cert ~/.docker/cert.pem --key ~/.docker/key.pem https://dockerhost:2376/images/json

I'm no go expert, but scanning the check_docker I see no way to provide a TLS certificate and key. Additionally, here is the result of a naive attempt:

$ /tmp/check_docker/check_docker-linux -base-url="https://dockerhost:2376"
CRITICAL: Get https://dockerhost:2376/v1.10/info: x509: certificate signed by unknown authority

Does check_docker need additional support added in order to connect to a TLS-secured docker daemon?

base-url flag doesn't work properly

Looks like the base-url flag isn't parsed properly making it impossible to override the default. Also, the program actually fails if the docker daemon isn't listening on the default endpoint (http://localhost:2375).

expected:

$ ./check_docker -base-url="https://$(boot2docker ip):2376"
CRITICAL: Get https://192.168.59.103:2376/containers/json: x509: certificate signed by unknown authority

actual:

$ ./check_docker -base-url="https://$(boot2docker ip):2376"
CRITICAL: cannot connect to Docker endpoint

newrelic / check_docker Goto Github PK

check_docker's People

Contributors

Stargazers

Watchers

Forkers

check_docker's Issues

No ability to check separate containers with same image name

May not work on AUFS backends

Connect to docker over unix domain socket

check does not timeout

Check returns incorrect Metaspace utilisation due to human readable units

Doesn't build with go 1.3 or 1.6 on Debian 8.3

https support

base-url flag doesn't work properly

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent