Additional rollup node metrics

Currently exposed metrics

• Level of the last inbox (single value)
• Number of messages (internal/external) in head's inbox
• Node info (version, commit hash, commit date) (single value)
• Rollup node info (rollup address, mode, genesis level, PVM kind) (single value)
• Processing time for each block (histogram)
  • history graph
  • time to fetch messages
  • time to execute PVM

Desirable metrics

Rollup node

Generic information

• LCC (local and on L1)
• Last published commitment (local and on L1)
• First available level (not GCed)
• Show full L2 head (/evm/blocks/current/number key in the durable storage, little-endian u256, to tell that the rollup continues to produce new blocks) (json, or table)
• Amount of transactions per layer 2 block (history graph)
  • TPS (current + ath)

GC

• Statistics about context GC (time, etc.) (history graph + histogram)
• Statistics about store GC

PVM

• Maybe provide a way to configure a "view" on durable storage to expose in grafana

Batcher

• Size of message queue (analogous to mempool of L1)
• Time/level of last batch
• Time of next batch?

Injector

(per injector)

• Size of queue (per operation tag)
• Size of tables injected/included
• Time of next injection round?
• nubmer of injected operations
• error rate

Generic performance

• CPU usage of machine
• RAM usage
• disk usage
  • data-dir
  • data-dir/storage (~ 100 MB)
  • data-dir/context (~ < 4GB)
  • data-dir/daily-logs
• file descriptors