Problem
lndmon keeps cumulative counters in memory, such as for total payments. On a pod restart they reset to zero.
This type of counter is not typical for prometheus exporters, but bridges a gap for embedded database.
Proposed fix
Optional flag, to, seed the counters from lnd's current state on startup (via the relevant RPCs recently became significantly faster on recent lnd), then resume incremental tracking. Across a restart, derived rate metrics should stay continuous instead of dropping to zero.
Acceptance
- After an lndmon restart, counters resume from approximately their pre-restart values (within tolerance) rather than from zero.
Problem
lndmon keeps cumulative counters in memory, such as for total payments. On a pod restart they reset to zero.
This type of counter is not typical for prometheus exporters, but bridges a gap for embedded database.
Proposed fix
Optional flag, to, seed the counters from lnd's current state on startup (via the relevant RPCs recently became significantly faster on recent lnd), then resume incremental tracking. Across a restart, derived rate metrics should stay continuous instead of dropping to zero.
Acceptance