Metrics Collection

[luca-moser]

This issue describes the different metrics we want to collect in GoShimmer and in the global network analysis server to assess network behavior and for debugging purposes.

Local metrics (GoShimmer node):

• Average lifetime of a peer connection (before getting dropped) (a)
• Virtual distance (min, max, mean) of connected peers (a)
• Size of the database (a)
• Inbound/outbound network traffic (bytes) (a)
  • FPC
  • Message gossiping
  • Autopeering
  • Analysis client
• Communication layer:
  • Messages count per message type (l)
  • Tips count (a)
• Value layer:
  • Value objects per second (l)
  • Tips count (l)
• Process:
  • CPU usage (l)
  • Memory consumption (l)
• FPC: (l)
  • Number of active conflicts
  • Number of finalized vote ctxs
  • Number of failed vote ctxs
  • Number of rounds it takes to finalize a vote context
  • Amount of queries received
  • Amount of replies NOT received
• Synced (l)

Network/Global metrics (analysis server):

• Prometheus:
  • Resource consumption (client):
    • Amount of cores
    • Amount of load/cpu usage
    • Amount of memory used/available
    • GOOS/GOARCH
  • Autopeering:
    • Amount of neighbors (histogram)
    • Network diameter
  • FPC:
    • Amount of conflicts
    • Avg. time it takes to finalize a conflict
    • Ratio between initial opinion
• Live dashboard:
  • FPC:
    • Current active conflicts with opinion per peer
  • Autopeering:
    • Avg. number of peers
    • Peers count
• Storage:
  • Finalized conflicts with the history for every node

[luca-moser]

Amazing