Loading feed…
Planet PostgreSQL0 Hot0 comments7 min read
All Your GUCs in a Row: max_logical_replication_workers
A deep technical breakdown of PostgreSQL's max_logical_replication_workers GUC explains that the pool it controls is shared cluster-wide across four kinds of workers: the launcher, leader apply workers (one per enabled subscription, permanent), table synchronization workers, and parallel apply workers (which linger idle after use rather than exiting). When the pool runs dry, nothing errors outright—subscriptions silently stall in an 'initializing' state, logs fill with repeated warnings, and retries are throttled by wal_retrieve_retry_interval, compounding delays across many tables. The piece walks through concrete reproductions on PostgreSQL 18.6, explains how max_worker_processes bounds the whole pool without cross-checking at startup, covers the special danger of setting the value to zero (a dangling replication slot pinning WAL and xmin on the publisher), and gives practical sizing guidance: count one slot per subscription permanently, one more per streaming=parallel subscription after first use, plus sync worker slots, plus the launcher, then round up max_worker_processes accordingly.
Source: https://postgr.es/p/9uK. 8 Sync News only summarizes and links out; content copyright belongs to the authors and original sources.
Recommended for you
Bình luận
Đang tải bình luận…
