perf_compare.sh: wait for CQL readiness on each Cassandra node

[mweiden]

Summary

Fix a startup race in start_cassandra_cluster where cassandra-stress could see Connection refused: /172.19.0.x:9042 warnings on peer nodes.

Root cause

The per-node readiness check after docker run was only wait_un_count $n (gossip view via nodetool status). UN ("Up Normal") flips true as soon as a node joins the gossip ring, but the CQL native transport on port 9042 binds several seconds later. When cassandra-stress connects to the seed and the driver opens connection pools to all peers it learned about via gossip, it races CQL startup on the just-joined peer and emits transport warnings.

The seed (cassA1) was already protected by an explicit cqlsh -e 'DESCRIBE CLUSTER' poll at line 110. The non-seed loop wasn't.

Fix

Add the same cqlsh -e 'DESCRIBE CLUSTER' poll inside the for n in 2 3 4 5 loop, after wait_un_count. Same pattern as the seed.

Test plan

• scripts/perf_compare.sh runs end-to-end without Connection refused warnings during the Cassandra phase
• All 5 nodes are CQL-ready before cassandra-stress starts
• perf-results/cassandra_*_t*.log files are populated as expected

🤖 Generated with Claude Code

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 0a77614b28

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".