mirror of
https://github.com/neondatabase/neon.git
synced 2026-05-30 19:40:39 +00:00
fc242afcc2
## Problem When testing tenant splits, I found that PG can get backpressure throttled indefinitely if the split is aborted afterwards. It turns out that each PageServer activates new shard separately even before the split is committed and they may start sending PageserverFeedback to PG directly. As a result, if the split is aborted, no one resets the pageserver feedback in PG, and thus PG will be backpressure throttled forever unless it's restarted manually. ## Summary of changes This PR fixes this problem by having `walprop_pg_process_safekeeper_feedback` simply ignore all pageserver feedback from unknown shards. The source of truth here is defined by the shard map, which is guaranteed to be reloaded only after the split is committed. Co-authored-by: Chen Luo <chen.luo@databricks.com>
neon extension consists of several parts:
shared preload library neon.so
-
implements storage manager API and network communications with remote page server.
-
walproposer: implements broadcast protocol between postgres and WAL safekeepers.
-
control plane connector: Captures updates to roles/databases using ProcessUtility_hook and sends them to the control ProcessUtility_hook.
-
remote extension server: Request compute_ctl to download extension files.
-
file_cache: Local file cache is used to temporary store relations pages in local file system for better performance.
-
relsize_cache: Relation size cache for better neon performance.
SQL functions in neon--*.sql
Utility functions to expose neon specific information to user and metrics collection. This extension is created in all databases in the cluster by default.