Deku performance metrics

d4hines commented 2 years ago

Keep in mind Prometheus collects timestamps on every sample and has a query language to do many transformations. Refer to its docs.

Things we want to measure with Prometheus:

[x] block rate
[x] transaction rate
[x] incoming message size/rate
[ ] how long does it take to broadcast a new block (should not be a lot, but it's a good metric to compare with the total time of a block, e.g. "of the total block rate X, we spend Y% hashing, signing, and serializing the new block")
[ ] mempool size for each node

d4hines commented 2 years ago

cc @lykimq @InfiniteSwerve

lykimq commented 2 years ago

where is the mempool define in Deku? is it the block_pool ?

ulrikstrid commented 2 years ago

Are the queries used documented somewhere? I want to add it to the grafana instance for our testnet

d4hines commented 2 years ago

This is being handled elsewhere.

marigold-dev / deku