Hi all, I wanted to bring attention to the recent proposal and pull request introducing the perfmon extension for Cloudberry:
Proposal: https://github.com/apache/incubator-cloudberry/discussions/1087 PR: https://github.com/apache/incubator-cloudberry/pull/1085 After reviewing the PR, it appears that the implementation is based on the legacy Greenplum gpperfmon monitoring framework. Files like gpmmon.c and others closely follow the original Greenplum design and code structure. This is understandable given the historical context of contributors, but this relationship is not currently disclosed in the proposal or the PR. I believe this should be made explicit for transparency and to clarify provenance and licensing considerations under the Apache Software Foundation. Additionally, the PR includes a file named README_hashdata.md, which appears to reference an unrelated internal effort ("hashdata-lightning"). It mentions various removed or disabled features such as gp_elog, check_disk_space, and gpperfmon_log_alert_level, but without context or explanation. This file seems inappropriate for an ASF-hosted repository and should either be cleaned up or removed from the PR. ------------------------------ Process concern: The PR was submitted before any discussion or feedback had taken place on the proposal thread. I’d like to suggest that we encourage a more deliberate process for future contributions: 1. Submit a proposal (or DISCUSS thread) 2. Allow time for community feedback and iteration 3. Open a PR once there’s some level of shared direction or consensus This becomes especially important for features derived from legacy systems or other projects, and for any contribution that introduces new architecture or visible user-facing components. ------------------------------ Suggestions: - The proposal should clearly state that the implementation is adapted from Greenplum gpperfmon. - The PR should remove README_hashdata.md unless it is rewritten to fit Cloudberry’s context. - Consider renaming components (e.g., from gpmon, gpperfmon) to cbmon, cbperfmon, etc., to align with Cloudberry’s branding and reduce legacy confusion. - Establish (even informally for now) a Proposal -> Discussion -> PR workflow for significant changes. This looks like a solid foundation for observability in Cloudberry, and I appreciate the initiative here. A little more transparency and process discipline will go a long way as the community grows. Best, -=e -- Ed Espino Apache Cloudberry (Incubating) & MADlib