feat: optionally reclaim expired account rent from operator #172
Conversation
| Ok((tda_accounts, pfda_accounts)) | ||
| } | ||
|
|
||
| async fn fetch_expired_claim_statuses( |
There was a problem hiding this comment.
This has an average of 30-60 seconds latency vs 700s+ for the existing approach for fetching claims by epoch (within the claims process) using get batched accounts, not counting sourcing the pubkeys from the very large merkle tree collection.
| #[allow(clippy::integer_division)] | ||
| #[allow(clippy::arithmetic_side_effects)] | ||
| #[allow(clippy::manual_div_ceil)] | ||
| pub fn pack_transactions( |
There was a problem hiding this comment.
This helps us achieve 200+ closed accounts per second on a single instance. We should re-use this where applicable. The claims process would be a great candidate.
There was a problem hiding this comment.
Claims can't be packed well just because the merkle proof is very long but might be able to get some improvements!
gzalz
changed the title
| // Use default timeout and commitment config for fetching the current epoch | ||
| let rpc_client = rpc_utils::new_rpc_client(rpc_url); | ||
| let current_epoch = rpc_client.get_epoch_info().await?.epoch; | ||
| (current_epoch - num_monitored_epochs)..current_epoch |
There was a problem hiding this comment.
Shouldn't this be offset by the number of epochs the claim status are active for? what are you using for num_monitored_epochs?
There was a problem hiding this comment.
We're filtering by the expires_at epoch in our getProgramAccounts calls so this should be correct. num_monitored_epochs will be the same as the CLI param.
The operator will claim and reclaim rent for claims that should be claimed in a monitored epoch or expire in a monitored epoch respectively.
| // Use default timeout and commitment config for fetching the current epoch | ||
| let rpc_client = rpc_utils::new_rpc_client(rpc_url); | ||
| let current_epoch = rpc_client.get_epoch_info().await?.epoch; | ||
| (current_epoch - num_monitored_epochs)..current_epoch |
|
Noticed the tip claim status filter was off by one byte. Addressed in latest commit. |
Problem
Our rent reclamation is currently an ad-hoc script. The tip router operator should handle operations like these without a need for manual action.
Solution