From 093763533820338249dd7ed054ea835fcf809e40 Mon Sep 17 00:00:00 2001 From: gubbins Date: Thu, 17 Aug 2023 13:32:06 +1000 Subject: [PATCH] Clarify use of read replicas The docs suggest that reads are evenly distributed between the master DB and the read replica DBs. Based on both the server code and the observed behaviour, I do not believe that is the case - I believe that (if any read replicas are available) read queries are distributed among the read replicas only. This is a useful distinction to keep in mind for site admins. --- source/scale/high-availability-cluster.rst | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/source/scale/high-availability-cluster.rst b/source/scale/high-availability-cluster.rst index 887db0834b2..6b76df8aee0 100644 --- a/source/scale/high-availability-cluster.rst +++ b/source/scale/high-availability-cluster.rst @@ -214,7 +214,7 @@ Migrating to NAS or S3 from local storage is beyond the scope of this document. Database configuration ^^^^^^^^^^^^^^^^^^^^^^ -Use the read replica feature to scale the database. The Mattermost server can be set up to use one master database and multiple read replica databases. Mattermost distributes read requests across all databases, and sends write requests to the master database, and those changes are then sent to update the read replicas. +Use the read replica feature to scale the database. The Mattermost server can be set up to use one master database and multiple read replica databases. Mattermost distributes read requests across all read replica databases, and sends write requests to the master database, and those changes are then sent to update the read replicas. On large deployments, consider using the search replica feature to isolate search queries onto one or more database servers. A search replica is similar to a read replica, but is used only for handling search queries.