maglub/nagios-cloudwatch

Nagios check for AWS Cloudwatch statistics/metrics

nagios-cloudwatch's Issues

Does not fetch metrics for AWS/ApplicationELB

No metrics can be fetched from ALB loadbalancers. Works fine with classics but not for ALB.

Tags to include in ./check_cloudwatch --ec2 and --elb

If a customer need to include certain tags in the output of a listing.

I.e. tags like Team, DeploymentState, etc.

Could be included in the config.yml file

Custom variables not defined in management packs

The management packs for op5 Monitor are not configured to prompt for the custom variables ($_HOSTEC2_ID$, $_HOSTEC2_REGION$) used in the check commands.

This results in a host wizard that does not ask for these IDs and therefore failing checks.

Some invalid threshold values in op5 management packs

The thresholds for the following service checks need to be set to 0, not 1 (they will never be triggered as the arguments are using soft threshold config, AGR+)

1. StatusCheckFailed_Instance (EC2)
2. StatusCheckFailed_System (EC2)
3. UnHealthyHostCount (ELB)

Not all instances in a region is listed with ./check_cloudwatch.rb --ec2

A customer reported that only 4 out of 6 configured and running instances were listed with:

./check_cloudwatch.rb --ec2

Add Billing monitoring

• Documentation on how to enable monitoring for billing in AWS
• Implement feature in check script
• Create check_command examples
• Document usage

Add "power state" as a check

Showing power state for an EC2 instance, with the possibility to monitor power state of the instance.

Billing monitoring - Impelement feature in check script - method for overall costs

• Create method for fetching overall cost

https://console.aws.amazon.com/cloudwatch/home?region=us-east-1#metrics:graph=!D03!E07!ET6!MN4!NS2!PD1!SS5!ST0!VA-PT12H~21600~AWS%25252FBilling~Currency~EstimatedCharges~Maximum~P0D~USD

http://docs.aws.amazon.com/AmazonCloudWatch/latest/DeveloperGuide/monitor_estimated_charges_with_cloudwatch.html#turning_on_billing_metrics

--powerstate on non existing instance ERROR: undefined method for nil:NilClass

root@monitor-test-app1:/opt/plugins/custom/nagios-cloudwatch# ./check_cloudwatch.rb --ec2 -i i-c9dd4c2d -C stage.yml --powerstate -c :1+
ERROR:   - undefined method `[]' for nil:NilClass

Adapt checkcommands.cfg so that it is easy to mix multiple regions

• Extra attribute, i.e. $ARG4$ in checkcommands.cfg
• Allow empty --region in script -> default to region in config file
• Should region be a custom variable for the host?

No specific error message for faulty keys

If the plugin is used with a pair of faulty keys it does not throw a describing error message like "Invalid access or secret key", but one that looks something like this:

ERROR: Could not get cloudwatch stats: NetworkOut
./check_cloudwatch.rb:838: undefined method `[]' for nil:NilClass (NoMethodError)

Billing monitoring - Impelement feature in check script - method for individual metrics

• Create method for fetching individual billing metrics

Support IAM Instance Roles

Would be great to be able to use the Role of the EC2 instance you're running on in order to avoid the need for credentials to be stored anywhere. Seems pretty simple to support - I've done a very quick and dirty local hack to accomplish this.

Add checkcommands.cfg entries to simplify multiple AWS accounts (config.xy.yml)

• Add config so that one simply can pass a path to a different config.yml file than default
• Add check in the script so that an empty --config can be passed (default to config.yml)

Billing monitoring - Documentation on how to enable in AWS

• Document how to enable the billing monitoring

References:

• http://docs.aws.amazon.com/AmazonCloudWatch/latest/DeveloperGuide/monitor_estimated_charges_with_cloudwatch.html
• http://docs.aws.amazon.com/AmazonCloudWatch/latest/DeveloperGuide/monitor_estimated_charges_with_cloudwatch.html#turning_on_billing_metrics
• https://kb.op5.com/display/GUI/Listview+filter+columns

Using the wrong config file for an instance does not show any warnings/errors

If the wrong config file is used (i.e wrong credentials to connect to collect metrics for a certain instance), it is not visible in OP5.

Test/adapt ELB monitoring when using 2 elb nodes in a configuration

Docu: Add filters for infrastructure and services

• Services: [services] host.groups >= "ec2-linux-std"
• Infrastructure: [services] host.name="EC2-infrastructure"
• Hosts: [hosts] groups >= "ec2-linux-std"

Billing monitoring - Impelement feature in check script - List billing metrics

• Create method for listing billing metrics

Adapt checkcommands.cfg so that it is easy to mix multiple regions

For example as an extra parameter $ARG3$ for --region, and that --region can be empty on the command line to default to config file.

Github Pages out of date

http://maglub.github.io/nagios-cloudwatch/

It appears that the versions listed on the Github pages site is not locked, causing the v2 of aws-sdk to install. Could that get updated? The repo itself is clear on version locking this, making it a non-issue if someone reads here first.

Timestamp inconsistency for --powerstate

The --powerstate outputs a different timestamp (different timezone) than the other metrics, since the script takes the timestamp from the local server, not from the AWS metrics. This since there are no timestamps delivered in the metric data.

Option 1: Check which timezone the instance resides in, and convert the time to this.
Option 2: Remove the timestamp, since the information is redundant. The info is shown in the "last checked" column in the web gui of most nagios installations.

Pass access and secret keys as arguments for the commands in the management pack

For portability and Nagios-ness I would recommend using custom variables for the access and secret keys. This would allow the user to enter these during the host wizard and remove the need to create a config.yml file on the host system.

It could also be great for people running managed environments where every instance might not be using the same AWS API keys for monitoring.

Bug: There is no --list-instnaces for --elb

[email protected]:/Users/malu/dev/nagios-cloudwatch $./check_cloudwatch.rb --elb --list-instances
Name: vpn-v001ec2 Id: i-91f62199 privateIp: 192.168.99.6 State: Zone: running
Name: kmg-test002ec2 Id: i-7ef62176 privateIp: 192.168.99.10 State: Zone: running
Name: win-test003ec2 Id: i-9956e691 privateIp: 192.168.99.20 State: Zone: stopped
Name: hilti-azb-v001ec2 Id: i-8e6bef86 privateIp: 10.5.2.10 State: Zone: running
Name: hilti-aza-v001ec2 Id: i-1b7bde10 privateIp: 10.5.1.10 State: Zone: running

--no-run-check does not seem to do anything

I was unable to see any difference in execution time of the script when the --no-run-check flag was used. If I enabled debugging I could still see something like this:

• Entering: EC2InstanceRunning
  • Checking running state of instance-id
  • Done checking running state of instance-id (running)

Is this feature currently implemented?

Monitor custom metrics

A user contacted me, asking for custom metric monitoring.

Hi,

Im implementing your cloudwatch plugin for our amazon environment, and im having some issues with collecting metrics from custom metrics I have added to my instances.

For instance amazon has not added RAM usage, but this is something that I have added, and can see in cloudwatch.

If I take one of those instances, and --list-metrics from it, that metric is not available.

Have you done anything like this before with custom metrics?

Regards,

Henrik Hjelm

He added the following link in a follow up email:

http://docs.aws.amazon.com/AmazonCloudWatch/latest/DeveloperGuide/mon-scripts-perl.html

Billing monitoring - Impelement feature in check script - Argument/Parameter

• Add parameter for choosing billing. I.e. --billing (in relation to --elb, --ec2)