This is an attempt to create an SNMP stats collector for various network devices. I grew tired of Cacti, I hate Zabbix, and I think that Collectd could do its job better. I also hate non-BULK SNMPWALKs.
Enter KOFTA — the piece of Rails code that was born with a single purpose: To replace Collectd and to feed InfluxDB with data, with style.
- KOFTA is back as of Christmas/New Years season 2017, this time with updated Web UI and less crappy code! I also fixed a nasty shameful bug in Sidekiq middleware I wrote for Kofta, now it should be less memory-hungry.
- As of March 2017, the app is being used succesfully for quite some time in 24/7 mode without any major hiccups (expect some occasional slowdowns when a huge network storm happens, or when my favourite Huawei router gets shy). Because of the great amount of my stupidity and laziness involved in the app code, Sidekiq processes tend to get fatter and fatter over time until they go swapping and anoter bad slowdown happens as a result. As a 'temporary hack' (haha, temporary!) I used Monit with a custom config that restarts Sidekiq workers when they exceed 900MiB RSS -- on my current environment and device/metrics set that results in restarting workers every 2 days or so. But because this setup uses 3 separate Sidekiq workers, there is no 'holes' in operating time.
- As of 2016-Aug-01, the app is almost fully functional. InfluxDB connection and data sending works well, Web UI is almost done, repeating parts are moved to layout and shared partials.
- Dynamic query schedule is implemented using
rufus-scheduler. Also, now it doesn't trigger jobs when Rails console is started, jobs start only when Rails App Server (or Puma) is up. - SNMP queries (in per-device workers) are parallelized using a thread pool. Derive metrics are handled internally, with Redis as intermediate data storage.
- Current repo also includes SMF manifests for Solaris/Illumos hosts.
TODO is moved to "TODO.org" file. Please refer to it from now on. If you want to look through what was in this section of the README, please use Git log.
This software is released under the MIT license.
- Ruby >=2.2 is required (2.3.x is recommended, JRuby would be ideal, but isn't tested yet);
- Sidekiq 4.x as ActiveJob queue manager, made by Mike Perham;
- Redis (required for Sidekiq, 3.0.3 or newer is best);
- Sqlite3 (internal DB);
- Docker (Optional, use Dockerfile and Docker Compose config provided in the repo);
- Ruby-SNMP gem for stat collection, made by Dave Halliday;
- InfluxDB::Client for DB connections, made and maintained by Todd Persen and Dominik Menke;
- Bulma CSS for Web UI, made by Jeremy Thomas.
git clone <repo_URL> <app_dir>;cd <app_dir>;bundle install;- To initialize database:
rake db:setup; This will init the DB with basic stuff (like, basic set of metrics that can be used to query devices) - Start Redis (use your OS service manager);
- Start Sidekiq with
bundle exec sidekiq; - If you want to test/develop it:
rails server; - If you want to see how it'll look/behave in real-er world:
RAILS_ENV=production rails server.