sensu-plugins / sensu-plugins-memory-checks Goto Github PK

I am getting following error will executing
[root@localhost bin]# /opt/sensu/embedded/bin/ruby check-memory.rb -w 700 -c 800

/opt/sensu/embedded/lib/ruby/gems/2.4.0/gems/sensu-plugins-memory-checks-3.0.0/bin/check-memory.sh: line 44: [: missing `]'

Its because there is no space on line 44
Instead of [ -z "$WARN"] || [ -z "$CRIT" ]; then it should be if [ -z "$WARN" ] || [ -z "$CRIT" ]; then

Your changelog mentions that the Ruby vmstat package has been removed as a dependency, but there are no instructions on how to install it in Sensu.

With the new 1.0 release, I'm getting this error:

/opt/sensu/embedded/lib/ruby/site_ruby/2.3.0/rubygems/core_ext/kernel_require.rb:55:in `require': cannot load such file -- vmstat (LoadError) from
/opt/sensu/embedded/lib/ruby/site_ruby/2.3.0/rubygems/core_ext/kernel_require.rb:55:in `require' from
/opt/sensu/embedded/lib/ruby/gems/2.3.0/gems/sensu-plugins-memory-checks-1.0.2/bin/check-ram.rb:36:in `<top (required)>' from /opt/sensu/embedded/bin/check-ram.rb:22:in `load' from
/opt/sensu/embedded/bin/check-ram.rb:22:in `<main>' Sensu::Plugin::CLI: ["Not implemented! You should override Sensu::Plugin::CLI#run."] 

I think setting a threshold on memory left vs memory used does not seem as natural as most plugins are based on used threshold.

If there is any chance of it passing I am more than happy to submit a PR. Either to adding additional doc as to how its used or changing it to be a usage threshold.

There was a major release that includes some major changes but it looks like it was never actually released. Can we get a release please? @eheydrick

@mattyjones @tas50 can we please have a tag and release of new gem?

#6

I didn't need to install the vmstat gem manually until now. Just installing this plugin worked by itself.

What changed ? Why can't the plugin installation pull the vmstat gem by itself if it is now required ?

There is no documentation of the currect usage of this plugin

It should be done as it is in check-memory-percent.sh. bc command is not installed by default on some systems. I've already rewritten it for my installation of Sensu and I can create pull requests.

Check Ram includes cache rather than checking memory in use

Following line override warning and critical threshold set when using "-u" or "--used" memory values.

L88 config[:warn] = 90
L89 config[:crit] = 95

If the current memory usage is at 91% and threshold as set as follows, it will not alert as threshold set on command line are overridden by above two lines

/opt/sensu/embedded/bin/ruby check-ram.rb --used -w 80 -c 90

Expected behaviour:

CheckRAM CRITICAL: 91% RAM used

Actual

CheckRAM WARNING: 91% RAM used

The is in reference to PR #11

@scalp42 @tas50

Thanks for fixing this. Filenames should never be hard coded into a script and should be passed in via ref. This should be changed.

Some files are named X-pcnt, X-percent, or X-percentage

There should be one way to name the files like this across the project or at least in here.

For example it is ok to have 20% of free ram but 40% of used ram is considered critical which is probably wrong.

Hi,
When client run on Centos 7 (systemd), the script metrics-memory-percent.rb returns wrong values

• run free command :

  free -b
  total used free shared buff/cache available
  Mem: 3880316928 424787968 145653760 33890304 3309875200 3145891840
  Swap: 0 0 0

• run script :

metrics-memory.rb
foo.memory.total 3880316928 1474542059
foo.memory.free 160362496 1474542059
foo.memory.buffers 0 1474542059
foo.memory.cached 387592192 1474542059
foo.memory.swapTotal 0 1474542059
foo.memory.swapFree 0 1474542059
foo.memory.dirty 40960 1474542059
foo.memory.swapUsed 0 1474542059
foo.memory.used 3719954432 1474542059
foo.memory.usedWOBuffersCaches 3332362240 1474542059
foo.memory.freeWOBuffersCaches 547954688 1474542059

I tried not using W FLAG but it sent me alert on 80 , so is there any way i can disable , some thing like -1

Hi,

I installed gem version 0.0.8. There is no bin/check-memory-pcnt.rb script. Instead, it is **bin/check-memory-percent.rb

the add_runtime_dependency for vmstat requires that the system this gem will be deployed to must have a gcc compiler in order to compile the vmstat binary. This is an annoying dependency, and is required even if the vmstat package is available on the system.

https://github.com/sensu-plugins/sensu-plugins-memory-checks/blob/master/sensu-plugins-memory-checks.gemspec#L44

the gem 'vmstat' is a runtime dependency:
https://github.com/sensu-plugins/sensu-plugins-memory-checks/blob/master/bin/check-ram.rb#L73

but is only configured as development dependency:
https://github.com/sensu-plugins/sensu-plugins-memory-checks/blob/master/sensu-plugins-memory-checks.gemspec#L48

As other check plugins within the repo have defaults which will send an alert. There should be a default for memory-check.sh or an error which prevents the check running without defined thresholds. The lack of thresholds will continue to show a healthy status until the system falls over. Even an arbitrary Threshold which will raise an alert seems better than none.

It appears that under the config[:megabytes], the math for free_ram, used_ram, and total_ram is actually dividing by 1 instead of the 1024's as intended. Here's the change I made that seemed to work:

 free_ram /= 1024 * 1024
 used_ram /= 1024 * 1024
 total_ram /= 1024 * 1024

The README currently has the following for a memory check resource definition example:

---
type: CheckConfig
spec:
  command: "check-memory.rb"
  handlers: []
  high_flap_threshold: 0
  interval: 10
  low_flap_threshold: 0
  publish: true
  runtime_assets:
  - sensu-plugins/sensu-plugins-memory-checks
  - sensu/sensu-ruby-runtime
  subscriptions:
  - linux

This isn't valid for a few reasons:

• No metadata, so no check name can be set.
• Missing the api_version attribute.
• The memory check command throws an error that both warning and critical need to be set.
  The following worked for me:

---
type: CheckConfig
api_version: core/v2
metadata:
  name: check-memory
  namespace: default
spec:
  command: "check-memory.rb -w 80 -c 90"
  high_flap_threshold: 0
  interval: 10
  low_flap_threshold: 0
  publish: true
  runtime_assets:
  - sensu-plugins/sensu-plugins-memory-checks
  - sensu/sensu-ruby-runtime
  subscriptions:
  - linux

Per the readme, check-ram.rb requires vmstat but that gem is not part of this gem's runtime requirements (not installed by default).

In order to make check-ram.rb functional for those consuming releases as Sensu Go asset builds, we need to explicitly install the vmstat gem as part of the build process.

I have a system which launches VMs and installs this sensu plugin, as well, as several others. I haven't launched any VMs for a while (2 or 3 weeks) and I started launching some yesterday, and noticed that on each machine check-ram.rb produces this stack trace:

Sensu::Plugin::CLI: ["Not implemented! You should override Sensu::Plugin::CLI#run."]
/usr/lib/ruby/1.9.1/rubygems/custom_require.rb:36:in `require': cannot load such file -- vmstat (LoadError)
    from /usr/lib/ruby/1.9.1/rubygems/custom_require.rb:36:in `require'
    from /var/lib/gems/1.9.1/gems/sensu-plugins-memory-checks-1.0.2/bin/check-ram.rb:36:in `<top (required)>'
    from /usr/local/bin/check-ram.rb:23:in `load'
    from /usr/local/bin/check-ram.rb:23:in `<main>'

The process which provisions these VMs will install these plugins from scratch on each VM so I imagine that the version running right now is not the same as it was a few weeks ago. The base image that the VM uses was updated but I don't think there was anything ruby/sensu-related in that set of updates.

The metrics-memory.rb plugin seems to be working fine, as well as all the sensu plugins.

Is there anything else I need to install/update/tweak to get this working again?

Hi,

My sensu-plugins gem was updated to latest version 1.3.0, making this plugins fail.
Could it be updated to support the updated dependency ?

/opt/sensu/embedded/lib/ruby/site_ruby/2.3.0/rubygems/dependency.rb:310:into_specs': Could not find 'sensu-plugin' (= 1.2.0) - did find: sensu-plugin-1.3.0
Checked in 'GEM_PATH=/opt/sensu/embedded/lib/ruby/gems/2.3.0:/opt/sensu/.gem/ruby/2.3.0', execute gem env for more information
from /opt/sensu/embedded/lib/ruby/site_ruby/2.3.0/rubygems/specification.rb:1439:in block in activate_dependencies' from /opt/sensu/embedded/lib/ruby/site_ruby/2.3.0/rubygems/specification.rb:1428:ineach'
from /opt/sensu/embedded/lib/ruby/site_ruby/2.3.0/rubygems/specification.rb:1428:in activate_dependencies' from /opt/sensu/embedded/lib/ruby/site_ruby/2.3.0/rubygems/specification.rb:1410:inactivate'
from /opt/sensu/embedded/lib/ruby/site_ruby/2.3.0/rubygems.rb:279:in block in activate_bin_path' from /opt/sensu/embedded/lib/ruby/site_ruby/2.3.0/rubygems.rb:279:insynchronize'
from /opt/sensu/embedded/lib/ruby/site_ruby/2.3.0/rubygems.rb:279:in activate_bin_path' from /opt/sensu/embedded/bin/check-memory.rb:22:in

sensu-install -p vmstats does not work

I have to use gem instead for vmstat

FWIW, I am using both Ubuntu 16/18 and CentOS 7 systems.

So near as I can tell this issues is caused by an incorrect parsing of the values returned by the free -m command as well as not calculating the free memory correctly.

Using the following output from the free -m command as an example,

                total        used        free      shared  buff/cache   available
     Mem:        3951        2915         224          32         812         731
     Swap:       1951         404        1547

The following code does not make much sense:

elif [ $os = "Linux" ]; then
  # Get total memory available on machine
  TotalMem=$(free -m | grep Mem | awk '{ print $2 }')
  # Determine amount of free memory on the machine
  set -o pipefail
  FreeMem=$(free -m | grep buffers/cache | awk '{ print $4 }')
  if [ $? -ne 0 ]; then
    FreeMem=$(free -m | grep Mem | awk '{ print $7 }')
  fi
fi

The first issue is where the code attempts to set the value for FreeMem.

This command:

FreeMem=$(free -m | grep buffers/cache | awk '{ print $4 }')
# does not actually return anything which causes the program bail out to the line after the "if [ $? -ne 0 ]; then" statement

# It probably ought to  be:
FreeMem$(free -m | grep Mem | awk '{ print $6 }')

However, this is also an issue as is it not the full picture of the free memory on the system.

FreeMem should be defined as:

Free Memory = Free + Buffers + Cached

So, using the the example data above we would have
Free Memory = Free (224 MB) + Buffers & Cache (812) = 1,036 MB.

A possible solution might look something like this:

elif [ $os = "Linux" ]; then`
  # Get total memory available on machine
  TotalMem=$(free -m | grep Mem | awk '{ print $2 }')
  # Determine amount of free memory on the machine
  set -o pipefail
  FreeMem=$(free -m | grep Mem | awk '{ print $4 }')
  BuffCache=$(free -m | grep Mem | awk '{ print $6 }')
 if [ $? -ne 0 ]; then
   UsedMem=$(free -m | grep Mem | awk '{ print $3 }')
 fi
fi

# Get percentage of free memory
TotalFree=$(($FreeMem+$BuffCache))
FreePer=$(awk -v total="$TotalMem" -v free="$TotalFree" 'BEGIN { printf("%-10f\n", (free / total) * 100) }' | cut -d. -f1)

Thank you for your effort on this project,
-Thomas

EDIT BY @majormoses: used triple backticks to form code blocks so its easier to read

Can anyone tell what units these metrics are in? For instance the '4356' from root used and the number after it.

[mesocool@blueserver bin]$ sudo ./metrics-disk-usage.rb
myserver.laptop.disk_usage.root.used 4356 1490035728
myserver.laptop.disk_usage.root.avail 25415 1490035728
myserver.laptop.disk_usage.root.used_percentage 15 1490035728
myserver.laptop.disk_usage.root.boot.used 221 1490035728
myserver.laptop.disk_usage.root.boot.avail 1693 1490035728
myserver.laptop.disk_usage.root.boot.used_percentage 12 1490035728
myserver.laptop.disk_usage.root.tmp.used 3 1490035728
myserver.laptop.disk_usage.root.tmp.avail 1849 1490035728
myserver.laptop.disk_usage.root.tmp.used_percentage 1 1490035728
myserver.laptop.disk_usage.root.var.used 722 1490035728
myserver.laptop.disk_usage.root.var.avail 2953 1490035728
myserver.laptop.disk_usage.root.var.used_percentage 20 1490035728
myserver.laptop.disk_usage.root.var.log.used 184 1490035728
myserver.laptop.disk_usage.root.var.log.avail 9047 1490035728
myserver.laptop.disk_usage.root.var.log.used_percentage 2 1490035728
myserver.laptop.disk_usage.root.var.log.audit.used 20 1490035728
myserver.laptop.disk_usage.root.var.log.audit.avail 3654 1490035728
myserver.laptop.disk_usage.root.var.log.audit.used_percentage 1 1490035728
myserver.laptop.disk_usage.root.disks.data304.used 46439 1490035728
myserver.laptop.disk_usage.root.disks.data304.avail 4253943 1490035728
myserver.laptop.disk_usage.root.disks.data304.used_percentage 2 1490035728
myserver.laptop.disk_usage.root.data.starfs1.used 197290 1490035728
myserver.laptop.disk_usage.root.data.starfs1.avail 705438 1490035728
myserver.laptop.disk_usage.root.data.starfs1.used_percentage 22 1490035728
myserver.laptop.disk_usage.root.data.fo.used 103321 1490035728
myserver.laptop.disk_usage.root.data.fo.avail 4795028 1490035728
myserver.laptop.disk_usage.root.data.fo.used_percentage 3 1490035728

Hi,

there is one serious problem with check-ram.rb:

`free -m`.split("\n").drop(1).each do |line|
free_ram = line.split[3].to_i if line =~ /^-\/\+ buffers\/cache:/ # rubocop:disable RegexpLiteral

It is not best way to relate on free -m. Why? Output can be different on different distributions. f.e.
Centos 6:

[root@centos6]# free -m
             total       used       free     shared    buffers     cached
Mem:          8192       1191       7000          2          0        953
-/+ buffers/cache:        237       7954
Swap:         4096          0       4096

Centos 7:

[root@centos7]# free -m
              total        used        free      shared  buff/cache   available
Mem:           8192         429         956           3        6806        7633
Swap:          4096         198        3897

This is causing false-positives in monitoring, because calling script gives output:

[root@centos7]# /opt/sensu/embedded/bin/check-ram.rb
CheckRAM CRITICAL: 0% free RAM left

I think it will be better to use /proc/meminfo - there is very small chance to change of file format.

Install vmstat gem into the path that sensu gems are expected to run from:
sudo /opt/sensu/embedded/bin/gem install vmstat --no-rdoc --no-ri

following above instruction, what is the expected destination this would install the gem file to?

I am able to install this gem as user "sensu", and gem installation destination is

/opt/sensu/.gem/ruby/2.3.0/gems

But rest of sensu gem dependencies are at

/opt/sensu/embedded/lib/ruby/gems/2.3.0/gems/

I am not too happy about dependencies spread across multiple locations...

Shell scripts should be rewritten to Pure Ruby in a way that is cross platform if available.

[ ] check-mem.sh
[ ] check-memory-pcnt.sh
[ ] check-swap-percentage.sh
[ ] check-swap.sh

sudo -u sensu /etc/sensu/plugins/check-memory-percent.rb
MEM CRITICAL - system memory usage: 27%

Is seems that there are no default tresholds, e.g. -w 80 and -c 90

Add compatibility with os x system with Darwin kernel

Hi, Could MacOS support get added to this since Sensu also has a Mac client now?

Ubuntu 16.04 has the RedHat style free output now.
It no longer has the buffers/cache section.