Release 1.10.22 Regression

[luzianscherrer]

Since release 1.10.22 has been marked with a regression issue, could you please issue a quick statement for the following points:

• What is the recommended course of action when one has already upgraded?
• Would a downgrade to 1.10.21 be safe?
• How can one determine if the instance is affected by the issue?

[karalabe]

You can downgrade, yes. No other ideas for now.

[cybersi]

Happened to me this morning. Decided to downgrade to v1.10.21, deleted the database and resynced. Is there a way to repair the local state if the corruption happens?

[MariusVanDerWijden]

Yes, you might be able to setHead to a block before you upgraded the node

[c0deright]

Do i need to issue the setHead command with v1.0.22 still running or do I have to upgrade to v1.0.23 first?

[karalabe]

Upgrade first

[c0deright]

Just for clarification: The setHead command will roll back the local blockchain and the state to the specified block as if geth didn't even see or know about the existence of all the blocks after that block that geth knew about right before the setHead command, right? Geth will then sync the "new" blocks as if geth was stopped for some days and needs to catch up to the chain head.
Is that correct?

[karalabe]

Exactly

[c0deright]

The question is: Are these errors
a) expected during resync after setHead and harmless and will the node be in a consistent state after full sync is done or
b) a sign of corruption that will still be around after resync of the rewinded blocks?

[karalabe]

If after setHead you are seeing missing trie node errors, then the db is fried beyond recovery and you shoudl resync (keep the ancients, delete all other files inside chaindata).