huaxingao commented on PR #3803: URL: https://github.com/apache/polaris/pull/3803#issuecomment-3988780168
> However, the response bodies can be easily large or even huge... Thanks @snazy for your comment. I agree that we can’t safely store/replay full response bodies (OOM risk with InMemoryIdempotencyStore, heavy DB/write amplification, and it’s not compatible with async/streaming responses). In my proposal, the intent is to persist only what’s needed to reproduce an equivalent response for later duplicates: ``` status (HTTP code) response_headers (small whitelisted set, e.g., Location, ETag, Content‑Type) response_summary (minimal body; e.g., metadata‑location, identifiers, error subtype) ``` Since this PR is already large and I’m splitting it into smaller PRs, I’ll follow up to (1) enforce a hard size bound / “not replayable” marker for large or streaming bodies (especially for in-memory), and (2) ensure heartbeat/lease updates don’t load/serialize response bodies (only touch required columns / projections). -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
