Graylog Cluster with MongoDB 8

Graylog Central (peer support)
,
1

1. Describe your incident:
I am building a Graylog Cluster, consisting of 2 Graylog Server and 2 Datanode Server.
As I followed the Installation Manual I build a MongoDB ReplicaSet named “rs0”.
Find the exact environment build below in Paragraph II.

While testing the loss of one Node (Graylog Server with MongoDB) I still get to the Authentication FrontEnd but after entering my credentials nothing really happens.

Starting the MongoDB Service on this ‘Maintenanced Node’ will fix the issue instantly.

2. Describe your environment:

  • OS Information: Debian 13

  • Package Version: Graylog Server 7.1.1. + Graylog Datanode 7.1.1

  • Setup:
    2 Graylog Server (with MongoDB ReplicaSet) + 1 Extra VM just for MongoDB Arbiter (providing Cluster Majority after loss of one Node)
    2 Graylog Datanodes (config mongodb_uri = mongodb://graylog01:27017,graylog02:27017,graylog03:27017/graylog?replicaSet=rs0) identical to server.conf on Graylog Server
    1 extra VM just for MongoDB Arbiter purposes (providing Cluster Majority after loss of one Node)

3. What steps have you already taken to try and solve the problem?

Testing Clusternode loss on PRIMARY / SECONDARY = same result.
rs.status() shows Arbiter is healthy as well as Replica Servers

4. How can the community help?

Do you ran into the same issue?
What should I edit / optimize to have a working Failover-Cluster?

Thanks for any helpful replies.

2

Hey @DoubleM,

I’m not quite sure I’m following what is going wrong here. To confirm a replica set should be comprised of 3 nodes that are able to vote and thus maintain quorum. Your build of primary, secondary and arbiter should be fine, assuming the Mongo service is active across all three nodes at the point one of them fails.

3

Hi @Wine_Merchant,
thanks for your advice. You gave the hint of having a 3 Node Mongo Cluster.
I followed that by removing the Arbiter Node and made it a full Member of this ReplicaSet.
I also added the new Host to the configs on Graylog Server and Graylog Datanode.
After testing - hurray - it worked as it should.
So I wasn’t aware, that this cluster has to contain of 3 full member Nodes.
As I always understood the cluster, it could be run with 2 full Membernodes and 1 additional Arbiter to keep the majority of Clustervotes when one Memberserver is down.

Thanks and kudos to you