Internal documentation only

This page has been marked as draft.

"no_majority" between collectors

Problem Copied

The collectors have stopped functioning as expected - an example being, no host or service checks are running

Error within the “/var/log/” logging Copied

[ERR] Failed to set security on database: bytearray(b'{"error":"error","reason":"no_majority"}\n')

An error generated by Opsview log file (opsview.log) due to having even collectors.

Implications Copied

When a cluster have even collectors, a change from one collector could affect the rest of its members.

• Collector Offline - if half of collector is down, the rest follows.
• Changes in Configuration - configurations on one node will automatically be applied to the other nodes.

Possible Cause(s) Copied

• ***Root Cause 1:*The collector cluster containing an even number of collectors; causing “Split Brain” issues

Possible Solution(s) Copied

• ***Solution Root Cause 1:*Best practice is to have uneven collectors per cluster

Steps Copied

• Check clustering evidence: Opsview - Collector Node Sanity Checks – Support - ITRS Group
• If the cluster has even collectors, the best practice is to deploy another collector in the cluster. This avoids split-brain issues and creates quorum

Definitions Copied

• Split-brain - hosts working independently and is not synced with other collectors. Collectors within a cluster should work as one as they are part of a group.
• Quorum - ensures majority of hosts actively communicate and decide for themselves.

Related Articles Copied

• Cluster and Collector Management - Managing clusters and collectors (itrsgroup.com)