feat!: simple stupid pruning #39
Conversation
There was a problem hiding this comment.
If I understand right, this is pruning everything before the last valset update, right?
If so, then I suggest keeping the name LastUnbondingNonce and prune before it. Because a valset is created also in other scenarios, other than a validator unbonding. So, it's better to keep attestations up to the last unbonding nonce
There was a problem hiding this comment.
If I understand right, this is pruning everything before the last valset update, right?
If so, then I suggest keeping the name LastUnbondingNonce and prune before it. Because a valset is created also in other scenarios, other than a validator unbonding. So, it's better to keep attestations up to the last unbonding nonce
|
to answer the question @sweexordious , this is pruning every attestation from 220,000 blocks ago, unless there is only a single validator set in the store, in which case it will not prune. closing this, as I think we will go with a different approach, however @sweexordious the test here might be something that we could emulate in whatever approach we end up going with |
Description
this PR attempts to simplify pruning by pruning attestations that are older than a threshold of blocks. It will stop pruning if there is only a single validator set update currently in the state, and will wait until there is another added before resuming.
I was too tired to write an integration test, but will tmrw