Prometheus Exporter Plugin

[stephenpdeos]

Supporting Documentation:

• Prometheus Plugin Architecture

Related Tasks:

• Enhancement: add prometheus exporter plugin that exposes key nodeos metrics mandel#67
• Enhancement: add prometheus exporter plugin that exposes key nodeos metrics AntelopeIO/leap#108

See #87

[bhazzard]

Items to include:

Node Configs

• Stuff from get_info

Node Status

• Unapplied transaction queue (Gauge)
  • number of transactions
  • byte size
  • speculative CPU time
• Blacklisted transactions (Counter)
  • by account
  • by key count
  • by contract count
• Subjective billing (Counter)
  • Number of accounts with subjective billing applied to them
• Scheduled transaction size (Gauge)
  • number of transactions
  • NOTE:
    • these are billed in RAM
    • relates to deferred transactions, which are deprecated
• Number of forks (Counter)
• Number of unapplied blocks (Counter)
• Number of dropped blocks (Counter)
• Number of transactions received by Peer (Counter)
  • Would include duplicates
  • Two parts: Peer could be P2P or an API
• Number of unique transactions received (Counter)
• Number of new connections (Gauge)
• Number of connections on wrong chain (Gauge)
• Uptime (Counter)
• CPU Usage by thread (Gauge)
• Disk space used by volume (blocks, ship, state, trace) (Gauge)
• Disk space available (Gauge)
• Current Block Number (Counter)

Other Prometheus Exporter Requirements

• Ability to expire metrics - clear on connection close

[matthewdarwin]

Can we please also add

• head_block_drift (gauge) current time - block time
• lib_block_drift (gauge) current time - lib time

[matthewdarwin]

Note "Disk space available" is also by volume.

[matthewdarwin]

For number of connections, consider inbound connections vs outbound connections.

• Total number of inbound connections.
• Total number of successful outbound connections.
• Total number of failed outbound connections (failure due to duplicate should be ignored... eg when outbound connection and inbound connection match each other). Failure could be to DNS failure, Wrong Chain, etc...

[bhazzard]

Add # of http transactions in progress