Go-ethereum: fatal error: flush stateDb: leveldb/table: corruption on table-footer

[carver]

Seeing output:

I0515 20:29:35.307528    5423 downloader.go:145] Block synchronisation started
I0515 20:29:46.580546    5423 chain_manager.go:564] INVALID block #333661 (09d2d6073181a027eb9bb88b196595c81b84bd524d046a2f1ae58829597e59f6)
I0515 20:29:46.580599    5423 chain_manager.go:565] invalid merkle root. received=c65b9384f74668976c77b7317ce583c36d55a857942b75f0810021cac8793119 got=65676f24f9b28d1de4247247a9f93c7bda3564db42baf17e91eb0700e525fd2d
I0515 20:29:46.580617    5423 sync.go:73] Block insertion failed: invalid merkle root. received=c65b9384f74668976c77b7317ce583c36d55a857942b75f0810021cac8793119 got=65676f24f9b28d1de4247247a9f93c7bda3564db42baf17e91eb0700e525fd2d
I0515 20:29:46.580653    5423 sync.go:112] Synchronisation failed: block downloading cancelled (requested)
I0515 20:30:10.144221    5423 sync.go:112] Synchronisation failed: peer's unknown or unhealthy
F0515 20:30:25.307284    5423 backend.go:475] fatal error: flush stateDb: leveldb/table: corruption on table-footer (pos=89792): bad magic number [file=008608.ldb] (Restart your node. We are aware of this issue)

Unfortunately, when I restart geth, I get exactly the same message again.

I got the message the first time on 9.20, and upgraded to 9.21. I've already run the upgradedb command when going from 9.19 to 9.20.

[obscuren]

You upgraded previously from 0.9.19(?) and you need to run geth upgradedb or delete your blockchain and redownload the chain

[carver]

upgradedb didn't work for me, but blowing away ~/.ethereum and starting over worked. There's still probably a bug somewhere, but I can't reproduce it, so I'm closing.

[obscuren]

That's correct. Old versions didn't properly delete all the database and left garbage in the stateDb. This has been fixed on newer versions of geth