AWS AmazonRDS documentation change
Summary
Added critical warning about max_slot_wal_keep_size configuration to prevent replication slot invalidation during integration downtime
Security assessment
Addresses operational reliability of replication, not security vulnerabilities. Explains data continuity risks from slot invalidation but doesn't mention security exploits or vulnerabilities.
Diff
diff --git a/AmazonRDS/latest/UserGuide/zero-etl.setting-up.md b/AmazonRDS/latest/UserGuide/zero-etl.setting-up.md index 17441155f..17b3b2635 100644 --- a//AmazonRDS/latest/UserGuide/zero-etl.setting-up.md +++ b//AmazonRDS/latest/UserGuide/zero-etl.setting-up.md @@ -67,0 +68,2 @@ Some of these parameters (such as `binlog_format`) are dynamic, meaning you can + * `max_slot_wal_keep_size = -1` (recommended), or sized large enough to cover peak WAL generation while the integration is not actively consuming. + @@ -78,0 +81,4 @@ Setting replica identity to log full rows [increases your WAL volume](https://ww +###### Important + +If `max_slot_wal_keep_size` is set to a finite value and the integration's logical replication slot accumulates WAL beyond that size while the integration is not actively consuming (for example, during integration creation, modification, or transient lag), PostgreSQL invalidates the slot. Once invalidated, the integration cannot resume replication from the source database. The default PostgreSQL value `-1` (unlimited) prevents this. If you need to bound WAL retention, choose a size that accommodates your database's write throughput during periods when the integration may not be actively consuming. +