From 06e96ae4c39ef782c6ad958be55b7b9252c5b8d8 Mon Sep 17 00:00:00 2001 From: Eoin Houlihan Date: Wed, 1 Feb 2017 23:03:56 +0000 Subject: [PATCH] Add replication doc --- docs/replication.org | 14 ++++++++++++++ 1 file changed, 14 insertions(+) diff --git a/docs/replication.org b/docs/replication.org index cbb57b1..86611fe 100644 --- a/docs/replication.org +++ b/docs/replication.org @@ -1 +1,15 @@ * Replication Across File Nodes +Replication of documents occurs when a file write happens to a primary. The +primary looks up the other nodes registered in the system. It then chooses one +of them randomly to serve as a secondary witness to the write and tells it to +replicate the file contents. This happens on a lightweight asynchronous thread +so that the user can receive the response from the primary that the primary +write has happened while the replication occurs in the background between +services. + +In this scheme, the primary node now has failover should it ever go down. We +have ensured that writes that occur on a primary have at least 1 other copy and +that the copy is placed on another node that is also visible to directory +servers within the system. This is important as we do not want replication +across nodes that are no longer reachable by directory services due to crashes +or network partitioning.