Skip to main content

Automatic retries of transfers

Overview​

A transfer can be automatically retried in Treasury (as of 25.7.2). If updating from a previous version, please update your default policy around the connector.

To understand what a retry really is, let's go over the possible states of a transfer or transaction:

  • preparing The transaction is waiting for inputs from the chain (e.g. gas fees, account sequence).
  • signing The transaction is waiting to be signed.
  • submitting The transaction is waiting to be broadcasted.
  • finalizing The transaction is waiting for confirmations.
  • succeeded The transaction has received enough confirmations to be considered final.
  • failed The transaction has failed and is no longer being watched.

A transfer is always associated with a transaction. By retrying a transfer, we create an additional transaction in the preparing state. This new transaction will get new inputs and be signed & broadcasted.

info

We also support "rechecking" a transfer, which moves it from either the succeeded or failed state to the finalizing state. This will cause Treasury to "recheck" if a transaction is on-chain or not.

Double sending​

A transfer should only move funds once, even if it can create multiple transactions. Treasury protects against double sending by checking that every transaction for a transfer spends the same resources. Perhaps they use the same nonce, or spend the same UTXO. Because of this native protection, users can take advantage of retrying transfers worry-free.

Creating an automatic retry policy​

While you are welcome to retry transfers automatically based on your own conditions, Treasury can do it autonomously as well.

Retry every 20 minutes for a specific error​

A retry policy is global for a specific chain. This will automatically retry SUI transfers once after 20m if the RPC error still contains "Transaction Expired".

update chain SUI { retry = { attempts = 2, after = "20m", message = "Transaction Expired" } }

Retry for multiple errors​

Match on multiple errors for a chain.

update chain SOL { retry = { attempts = 3, after = "20m", message = ["Attempt to debit an account but found no record of a prior credit", "Blockhash not found"] } }

Retry if not succeeded​

It's best to be specific about which RPC errors you want to retry on. Otherwise you can omit message to match anything.

update chain SOL { retry = { attempts = 10, after = "20m" } }