26.1 Antalya port - improvements for cluster requests#1414
26.1 Antalya port - improvements for cluster requests#1414ianton-ru wants to merge 14 commits intoantalya-26.1from
Conversation
…e_count_in_datalake 25.8 Antalya ports: Read optimization using Iceberg metadata
…wipes_my_snot 25.8 Antalya: Fixes for #1069
…el_segfailt Check multiplexed connection before usage
…_in_task Small iceberg read optimization
…orage_cluster_profile_events Profile events for task distribution in ObjectStorageCluster requests
…_storage_cluster_distribution
…n_unexpected_shutdown Fix segfault on unexpected node shutdown
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 9bd95cb8cd
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".
| replica_to_files_to_be_processed.erase(number_of_current_replica); | ||
| for (const auto & file : processed_file_list_ptr->second) |
There was a problem hiding this comment.
Keep task list before removing lost replica entry
In rescheduleTasksFromReplica, the code erases replica_to_files_to_be_processed[number_of_current_replica] and then immediately iterates processed_file_list_ptr->second. That iterator points to the erased element, so this is undefined behavior on the replica-failover path. When a replica drops mid-query, this can crash the coordinator or prevent tasks from being re-queued reliably.
Useful? React with 👍 / 👎.
| if (read_context.packet.type == Protocol::Server::Data) | ||
| read_context.has_data_packets = true; |
There was a problem hiding this comment.
Ignore empty Data packets when deciding task rerun
The failover guard marks has_data_packets true for any Protocol::Server::Data packet, but empty Data blocks are valid and do not imply rows were processed. If a replica sends only empty Data blocks and then disconnects, the EOF handler will skip the ConnectionLost rerun path (!has_data_packets is false), so unprocessed tasks are not reassigned even with skip_unavailable_shards enabled.
Useful? React with 👍 / 👎.
Changelog category (leave one):
Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):
Frontports for Antalya 26.1
CI/CD Options
Exclude tests:
Regression jobs to run: