Slow reorg on Fractal Bitcoin

[pokrovskyy]

Before you kick me for opening a ticket in main ord repo, please hear me out 😄

I submitted an issue in Fractal Bitcoin ord repo fork about reorg handling being beyond slow taking 5 hrs to roll back a 2-block-deep reorg. The reorg handling logic has NOT been changed in the Fractal Bitcoin ord fork, so I would assume this applies to main ord too. I maintain Bitcoin index as well and frankly I haven't seen a rollback on Bitcoin for quite a while, so can't comment on how quick is it there. And even if it takes time, it's so rare and can be tolerated. On Fractal Bitcoin it seems to happen every couple hrs paralyzing the index for 5 hrs which is unacceptable.

I would assume a 2-block reorg rollback should not take 5 hrs, even if the index is 380gb, or is this expected?

[so7ow]

Did you say Fractal? 🤣

[gmart7t2]

It's slower on fractal because the index is bigger. What I do is only index a block in ord once it is 3 blocks deep in the chain. Then you almost never get affected by reorgs.

[pokrovskyy]

Makes sense @gmart7t2! Is there any config setting for such indexer lag?

UPD: it doesn't seem like there was any such logic, added it to Updater.get_block_with_retries loop, will see how it goes