From 815a98264958aeca2607f2d98e0629c09d48d020 Mon Sep 17 00:00:00 2001 From: serprex <159546+serprex@users.noreply.github.com> Date: Fri, 2 Oct 2026 20:23:18 +0000 Subject: [PATCH] Remove experimental banner, qualify physical slot recommendation --- README.md | 11 ----------- docs/failover.md | 6 ++++-- docs/operations.md | 10 ++++++---- 3 files changed, 10 insertions(+), 17 deletions(-) diff --git a/README.md b/README.md index aaad8727..95905600 100644 --- a/README.md +++ b/README.md @@ -1,14 +1,3 @@ -# walshadow - -> [!WARNING] -> **Experimental** -> -> **Project status: Development Preview** -> -> walshadow is under active development and being hardened through real-world -> testing and feedback. We recommend validating it thoroughly with your -> workloads. Interfaces and behavior may evolve - walshadow replicates PostgreSQL rows into ClickHouse from physical WAL, including initial load, continuous changes, schema evolution, restarts, and planned source switchover diff --git a/docs/failover.md b/docs/failover.md index c80a0f29..e739c0d8 100644 --- a/docs/failover.md +++ b/docs/failover.md @@ -15,8 +15,10 @@ move DNS - required WAL remains available - when using slots, target already has physical slot covering resume position -Physical slots are not synchronized to standbys. Create target slot before -switchover with storage and WAL coverage appropriate for deployment +Physical slots are not necessary when using object storage, since archive +covers WAL retention. Physical slots are not synchronized to standbys. Create +target slot before switchover with storage and WAL coverage appropriate for +deployment ## Procedure diff --git a/docs/operations.md b/docs/operations.md index a175d7f9..ca6797d2 100644 --- a/docs/operations.md +++ b/docs/operations.md @@ -124,7 +124,9 @@ position. Reserve them for recovery drills or operator-directed rebuilds ## Retain source WAL -Configure a physical replication slot for routine deployments: +Configure a physical replication slot when not using object storage. With +object storage, archive already covers WAL retention and slots are not +necessary: ```sql SELECT pg_create_physical_replication_slot('walshadow'); @@ -143,9 +145,9 @@ immediately. `init` reports SQL above so slot can exist before daemon starts Live source moves never create target slot because a new slot cannot protect earlier resume position. Pre-create target slot before planned switchover -Without slot, ensure `wal_keep_size` or continuous archive covers worst-case -outage and backlog. Missing source WAL stops replication rather than skipping -data +Without slot, ensure object-storage archive or `wal_keep_size` covers +worst-case outage and backlog. Missing source WAL stops replication rather +than skipping data ## ClickHouse interruptions