1

We're running a parachain on Polkadot Relaychain. The parachain collator binary is built based on polkadot-sdk v0.9.29

When we try to upgrade binary to a new version which is built based on polkadot-sdk v1.1.0, it failed to launch and shows the following errors:

2024-05-09 05:27:26.036 INFO main sc_cli::runner: version 0.1.0-a1a4f683e67
2024-05-09 05:27:26.036 INFO main sc_cli::runner: by Parity Technologies <[email protected]>,2017-2024
2024-05-09 05:27:26.036 INFO main sc_cli::runner: Chain specification: xxxx
2024-05-09 05:27:26.036 INFO main sc_cli::runner: Node name: round-effect-1351
2024-05-09 05:27:26.036 INFO main sc_cli::runner: Role: AUTHORITY
2024-05-09 05:27:26.036 INFO main sc_cli::runner: Database: RocksDb at ...
2024-05-09 05:27:28.941 INFO main crust_collator::command: Parachain Account: ...
2024-05-09 05:27:28.941 INFO main crust_collator::command: Is collating: yes
2024-05-09 05:29:52.714 INFO main sub-libp2p: Local node identity is: ...

2024-05-09 05:29:52.735 INF0 main parachain::db: [Relaychain] Migrating parachains db from version 1 to version 2 ...
2024-05-09 05:29:52.919 INFO main parachain::db: [Relaychain] Migration complete!

2024-05-09 05:29:52.926 WARN tokio-runtime-worker sc service::builder: [Relaychain] The Networkstart returned as part of "build network’ hasbeen silently dropped
Eror: service(Application(Aplication(lo(custom { kind: other, error: Error { message: "Invalid argument: column families not opened: col5,o14,col3,col2,col1,colo"}}))))

If I purge the whole db and run the new collator binary, it works. But it would take a very long time to sync the whole Polkadot Relaychain db.

Any ideas on how to fix this problem?

3
  • By the way, you can use warp sync or a relay chain snapshot. This way, you won't have to wait long to synchronize the entire database.
    – AurevoirXavier
    Commented May 10 at 2:10
  • Thanks for pointing this out, I will give it a try!
    – Brian Wu
    Commented May 10 at 15:19
  • warp sync do work and it's really fast! Thanks!
    – Brian Wu
    Commented May 11 at 6:13

1 Answer 1

0

I only encounter this error when I accidentally kill the process. Then, the database becomes corrupted due to the unexpected exit. I suspect you've experienced the same.


By the way, how many collators do you have? Do they all experience this error? Because I don't think it's a Substrate upgrade issue.

6
  • I have four collators running. I tried to upgrade three of them, all these three nodes encounter the same error. Before I upgrade, they all work good. So I don't think this is data corrupted issue, it must be something inconsistent with the Substrate side.
    – Brian Wu
    Commented May 9 at 15:10
  • What if you downgrade the node and it starts working again?
    – AurevoirXavier
    Commented May 10 at 2:03
  • The db version has been upgraded from 1 to 2 (as you can see in the log of this topic), so when I downgrade the node, it would exit directly with the logs like 'db version from higher version node, cannot run.'
    – Brian Wu
    Commented May 10 at 10:15
  • Did you stop the node while it was upgrading the database? That might also cause this issue.
    – AurevoirXavier
    Commented May 10 at 10:40
  • No, I didn't stop it. The log shows 'Migration Complete' as you can see in the log of this topic. After it shows ‘Migration Complete', it errors out with the error message '...Invalid argument: column families not opened...'
    – Brian Wu
    Commented May 10 at 14:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.