Loading
[fix] TRON Stuck Outbounds
This contains a couple fixes for issues causing TRON stuck outbounds:
- Instead of using a truncated
time.Now()(which is non-deterministic near the boundary), we use the timestamp from thethornodeblock of the outbound height. - Reduce the ref block interval to 100 blocks (5 minutes) so inconsistent scanner/daemon heights do not exacerbate losing consensus on the outbound bytes to sign. The last 65536 blocks are valid for this field (https://developers.tron.network/docs/tron-protocol-transaction) so should not cause issue to use a slightly older one. We do not want it too long as nodes will need to wait for the next reference interval block at init before they have a valid reference block to join signing parties.
- Since we now use a deterministic timestamp from the
thornodeheight of the outbound transaction, we increase theTimestampValidityto 20 minutes (signing period is 30 minutes) from that timestamp, so it should be valid to broadcast for the first 20 minutes of the 30 minute signing period. - Treat broadcast error
DUP_TRANSACTION_ERRORas success. Otherwise a node can get temporarily stuck processing an already complete outbound, and not join consensus for subsequent ones.
Edited by Ursa (9R)