Grafsy

Grafsy

Release cycle

Go!

How does it work

• Goroutines (client, server, monitoring)
• Channels

• Github https://github.com/leoleovich/grafsy
• Jenkins stage job with testing
• Tags
• Jenkins job with tag version + build number (1.100.1-ig114)

• Fast!
• Python did not work out
• No dependencies on servers
• Fancy/New

Black box

What do we get?

• Service which is easily holding 2K RPS
• Reliable metrics delivery
• Easy maintenance of Graphite cluster
• Filtering bad metrics (less load on cluster)
• Sending interval (avoids peaks)
• Easy interface to communicate (localhost or file)
• Aggregation functions

Watchdog

Systemd

Debian package

• Grafsy calls sd_notify every N seconds
• If systemd did not get call - service will be restarted

• Universal supervisor for all Linux
• Nice interfaces to monitor status
• Dependencies
• ...
• Watchdog

• Dependencies
• Easy to check which files are managed by
• Easy to install previous version
• Postinst scripts
• ...
• Built by Jenkins