Sucks up metrics from various external sources and puts the data into internal systems. This is a good thing to do because whilst all the external SAAS services we use are fantastic, it's also sometimes very useful to be able to be able to view and manipulate the data in one place. For example, we might want to make a single graph with some internally generated application metrics, with some perfomance metrics from New Relic, with some metrics from an AWS Elastic Load Balancer. Perhaps. Anyway, vacuumetrix is a collection of ruby scripts that talk to various APIs and allows the collection of this data in one place.
- New Relic
- AWS Cloudwatch (some services)
- Neustar Web Performance Management
- facebook "likes"
- twitter "followers"
git clone https://github.com/99designs/vacuumetrix.git
apt-get install ruby build-essential libcurl3 libcurl3-gnutls libcurl4-openssl-dev
gem install json
If you're outputting to ganglia gmond you'll also need to
gem install gmetric
The config.rb file contains all the local configuration variables.
cd conf cp config.rb-sample config.rb
update config.rb with your organization's API key and account number
gem install curb xml-simple
update config.rb with your organization's AWSAccessKeyId and AWSSecretKey with permission to read the Cloudwatch API
gem install fog
update config.rb with your organization's API Key and Secret
gem install curb time
gem install twitter
The scripts are in the bin directory and are designed to be run from cron every minute. Generally if no argument is supplied to the script it'll grab all the metrics. If an argument is supplied it'll be more specific.
* * * * * /opt/vacuumetrix/bin/NewrelicEnduser.rb 123 metricyouwant
Note- The AWS CloudWatch scripts depending on the size of your infrastructure and options specified may take far longer than one minute to run. If you have several hundred instances, this is likely the case. Adjust the offset times and schedule less frequently. To adjust the offsets and scheduling time, just test the scripts with 'nc' or similar utilities to measure runtime. You are still able to maintain minute granularity but fetched less often on slightly longer intervals. Also just for clarity, GET requests currently cost $0.01USD per 1,000. If you have 500 instances and you are fetching 5k metrics every 5m, that is ~$430/month.
(Note: If you haven't already done so you will need to activate the New Relic REST API. See here: http://blog.newrelic.com/2011/06/20/new-data-api/ Get New Relic End User (RUM) stats. Supply two args, app and metric.
Get the threshhold values for all your applications. This includes average RAM, CPU, DB etc.
Get EC2 instance metrics. This includes average CPUUtilization, DiskReadBytes, DiskReadOps, DiskWriteBytes, DiskWriteOps, NetworkIn, and NetworkOut.
Get EBS metrics. No arguments. Run every 5 minutes. No point running more frequently.
Get Elastic Load Balancer metrics. Supply the name of the ELB or multiple ELBs and and optionally --start-offset and --end-offset (in seconds).
Get RDS metrics. Optionally supply the name of the Relational Database Service instance. (Tested with MySQL). YMMV.
Count the number of EC2 instances of each flavor. Also, capture instance tags if specified in config. No arguments.
Get Elasticache metrics. Interesting ones anyway. Supply the name of the Engine that can be "redis" or "memcache". (if use both need setup 2 cron lines)
Get DynamoDB metrics. Specify table_name as argument and optionally --start-offset and --end-offset (in seconds).
Get Neustar Web Performance Metrics. For each monitor get duration and status for each monitored location. No arguments. Run every 5 minutes. (Not very efficient at the moment. Needs some tuning).
Argument is the name of the page you want to check the like count of.
Argument is the name of the twitter user you want to check the number of followers of. You will need to create an application to generate the required OAuth keys here: https://dev.twitter.com/apps.
There is a bug discussed here: https://github.com/fog/fog/issues/2284 that causes the following error when specifying some CloudWatch metrics:
[excon][WARNING] Invalid Excon request keys: :host
This is fixed on master of Fog but has not been released yet; fix is here: https://github.com/fog/fog/commit/de07ac9016d00d385446820f6a945c7da5dc55b3
If you wish to manually patch it, it is a one-line removal.
- Other AWS metrics
- Google Analytics
- Retry class
- YML config.
Follow vacuumetrix on twitter for updates https://twitter.com/vacuumetrix Pull requests are appreciated.