@@ -15,14 +15,14 @@ Replica Set Arbiter
1515
1616.. start-content
1717
18- In some circumstances (such as you have a primary and a secondary but
19- cost constraints prohibit adding another secondary), you may choose to
20- add an arbiter to your replica set. An arbiter does **not** have a copy
21- of data set and **cannot** become a primary. However, an arbiter
22- participates in :ref:`elections for primary <replica-set-elections>`.
23- An arbiter has exactly ``1`` election vote.
18+ In some circumstances (such as when you have a primary and a secondary,
19+ but cost constraints prohibit adding another secondary), you may choose
20+ to add an arbiter to your replica set. An arbiter participates in
21+ :ref:`elections for primary <replica-set-elections>` but an arbiter does
22+ **not** have a copy of the data set and **cannot** become a primary.
2423
25- .. versionchanged:: 3.6
24+ An arbiter has exactly ``1`` election vote. By default an arbiter has
25+ priority ``0``.
2626
2727 .. include:: /includes/fact-arbiter-priority.rst
2828
@@ -69,6 +69,35 @@ Replica Set Protocol Version and Arbiter
6969
7070.. include:: /includes/extracts/arbiters-and-pvs-with-reference.rst
7171
72+ .. _rollbacks-multi-arbiters:
73+
74+ Concerns with Multiple Arbiters
75+ -------------------------------
76+
77+ Use a single arbiter to avoid problems with data consistency. Multiple
78+ arbiters prevent the reliable use of the majority write concern.
79+
80+ To ensure that a write will persist after the failure of a primary node,
81+ the majority write concern requires a majority of nodes to acknowledge
82+ a write operation. Arbiters do not store any data, but they do
83+ contribute to the number of nodes in a replica set. When a replica set
84+ has multiple arbiters it is less likely that a majority of data bearing
85+ nodes will be available after a node failure.
86+
87+ For more information, see this `Knowledge Base
88+ <https://kb.corp.mongodb.com/article/000019126/>`__ article on multiple
89+ arbiters.
90+
91+ .. warning::
92+
93+ If a secondary node falls behind the primary, and the cluster is
94+ :method:`reconfigured <rs.reconfig()>`, votes from multiple arbiters
95+ can elect the node that had fallen behind. The new primary will not
96+ have the unreplicated writes even though the writes could have been
97+ majority committed by the old configuration. The result is data
98+ loss.
99+
100+ To avoid this scenario, use at most a single arbiter.
72101
73102Security
74103--------
@@ -99,3 +128,17 @@ exchanges are not encrypted.
99128
100129As with all MongoDB components, run arbiters in trusted network
101130environments.
131+
132+ Example
133+ -------
134+
135+ .. start-content-even-votes-example
136+
137+ For example, in the following replica set with 2 data-bearing members
138+ (the primary and a secondary), an arbiter allows the set to have an odd
139+ number of votes to break a tie:
140+
141+ .. include:: /images/replica-set-primary-with-secondary-and-arbiter.rst
142+
143+ .. end-content-even-votes-example
144+
0 commit comments