Nagios Support Forum

Hi guys, is it possible to cluster two existing Nagios Log Server? Note that our servers are licensed. Thanks a lot

Unfortunately there isn't a good way to merge two existing clusters.

hmmm I found this guide:
Re-clustering Procedure:

1. Stop Elasticsearch
2. Edit /usr/local/nagioslogserver/var/cluster_uuid
2.1. Enter the cluster_ID of the Primary Nagios Log Server
3. /usr/local/nagioslogserver/var/cluster_hosts
3.1. Enter the IP address of the Primary Nagios log Server
4. Both servers must have the same node_uuid
5. Start Elasticsearch

is this possible?

Add a "step 1.5" to wipe all existing ElasticSearch data and cluster state, and it *MIGHT* work.

Trying to join 2 individual ElasticSearch nodes with individual data and cluster state into a 2-node cluster doesn't generally work out well. Particularly with each individual node likely sharing index names. It's really much easier to just start from scratch (by joining a "clean" node to one of your existing nodes) if you don't have data to preserve. If you do have data to preserve, you'd need to dump it and "re-send" it to the newly created ElasticSearch cluster. I'd think this could be accomplished via the snapshots and maintenance section by physically moving the Curator snapshots, though I am unsure of the level of official support for such work.

We don't have any official documented way to merge, but using snapshots may do the trick and be most straight forward. I'm not aware of anyone that's used them to merge so I'll look into it and do some testing on my end and let you know what we find.

oh thanks please let me know the results, though we don't need to keep the data right now since these is a new setup for production. The guy before me just installed them separately which is supposed to be clustered

Using the snapshot method didn't work out since it will complain about the other existing indices with the same name.

Since data retention isn't an issue, I would go the route of installing NLS on a new instance and then joining it to the other machine to form a cluster.

Yeah, I think that would be the best option to cluster