Ring SQA
Discovers NLNOG Ring nodes by monitoring /etc/hosts with inotify. UDP pings each node periodically recording latency as microseconds in SQL database
Currently 4 threads
- main thread, launches everything and finally gives control to Analyze class
- querier thread, sends queries and waits for responses, populates database
- responder thread, waits for queries and echoes them back
- inotify monitor thread
Use
ring-sqad --help ring-sqad --daemonize
Todo
- Querier loop should sleep dynamically between nodes to spread CPU/network demand
- Analyzer class should actually do something (use average of numbers before median as norm, if last Y measurements are Z times above norm (or more than X standard deviations?) raise alarm?