Details
-
Improvement
-
Resolution: Fixed
-
Major
-
4.1
-
None
Description
Adding the summary of the discussion shared by Chin:
From: Chin Hong <chin@couchbase.com>
Date: Thursday, March 24, 2016 at 12:51 PM
To: Amarantha <amarantha.kulkarni@couchbase.com>
Subject: FW: support contact
Amarantha:
We recommend that customers run with a min of 3 CB server instances in a production cluster. Support has a challenge in that when customers call in with issues when they are running with 1 or 2 CB instances in production and Support tells them that the min recommended configuration is 3 CB server instances, customers often complain that it is not well documented.
We need to improve the production deployment guidelines in our doc, and summarize the issues customers may need to consider when running with 1 or 2-node deployments. Here are the info from DaveF:
2 node cluster
No auto failover
Max of 1 replica
System needs to be at < 50 % load capacity (or when 1 nodes fails the other will too)
1 node “cluster” (actually not a cluster at all)
All the limitations of a 2 node cluster
Single node failure means recreating the cluster from scratch and restoring the data from backup (lengthy downtime)
Single node failure means more data loss than usual (no in-memory replicas)
Constant warning in the UI reminding you that you have no replicas
No failover (in addition to no auto failover)
No ability to add services dynamically (if you want to add query or indexing you need to unprovision and reprovision and dropping in the old datafiles)
No rolling upgrade; only offline
No rolling restart; only offline
Host name / IP must be set explicitly (not automatically picked up during clustering)