awaitSchemaAgreement not ignoring ghost node

[fesiqueira]

Please answer these questions before submitting your issue. Thanks!

What version of Cassandra are you using?

3.11.11

What version of Gocql are you using?

v1.6.0

What version of Go are you using?

1.21.6

What did you do?

Running migrations that use WaitForSchemaAgreement

What did you expect to see?

The migrations to finish successfully

What did you see instead?

gocql: cluster schema versions not consistent: [<uuid1> <uuid2>]

---

Describe your Cassandra cluster

I have a cluster of 50+ nodes. One of them was replaced a few days ago but the system.peers was left in an inconsistent state, where the replaced node is still showing in there with a different schema_version.

The nodetool status command prints out the "correct" output, where the old node is NOT shown.

The system.peers table shows the old node and all columns are filled with some value, like:

peer         | rpc_address  | schema_version
10.100.XX.XX | 10.100.XX.XX | <uuid2> 

The old node is unreachable.

[fesiqueira]

I can work on a fix for it, but I'm not sure what to do to fix it. I know that drivers for Python and Java check the node state when asserting if a node is a valid peer.

[fesiqueira]

Hi @joao-r-reis,

I wanted to check in on this PR/issue since it has been open for a while. Does it make sense to keep it open? If so, could you review it when you have a chance? If it's not relevant, I’m happy to close it.

Thanks for your time!

[joao-r-reis]

Yeah it makes sense to keep it open, right now we're prioritizing some other issues for the upcoming 2.0 release but this is definitely something that should be fixed at some point. Can you create a JIRA?

[fesiqueira]

Done! Opened CASSGO-52.