Skip to content
Jan Vilhuber edited this page Apr 15, 2015 · 15 revisions

Possible Performance Improvements

  • split off the StopToken from the DeviceInfo entry, and let the web-server directly manage it, so we can avoid the RPC call. Possibly even keep it in memory, but then we have to make sure any changes are properly sync'd with the DB. This would also allow us to respond to the client quicker, i.e. generate a stop token, respond and make a go-routine throw the request over to the RPC server.

TODO:

  • Rate limit incoming requests. Use some negroni middleware, preferably (see Utils/middleware.go)
  • Cleanup the structure cleanup, so that structures aren't cleaned up out from under a running goroutine.
  • do some log rotation for pinger and webserver logs
  • inspect code for resource leakage
  • inspect code for concurrent-access issues to data struct (related to the cleanup item above)
  • need something to email on crash or panic in logs. Perhaps that's what telemetry will be for.
    • supervisord events might be able to do this, if I log recoverCrash to stderr as well as the logfile.
    • or logwatch?
  • investigate telemetry for pinger
Clone this wiki locally