After implementation, there is only 1 primary master node (which performs cluster-wide tasks) and multiple member nodes.
Is traffic is distributed via load balancer to all the member nodes only, or also to primary node?
What happens when the primary node goes down?
Does one of the member nodes become the primary then? If yes, is there a pattern as to which one?
Answer to this question could be obvious, but just to confirm my understanding…
For effective HA, we need to have HA separately implemented for DB and the NFS storage also (apart from the Artifactory servers), right?
1. The primary node functions as a regular node and traffic will be distributed to it as well
2. The system can continue to operate as long as at least one of the Artifactory nodes is operational, without the primary node, any cluster-wide tasks such as replication and GC will not be run.
The selection of a new primary node is manual; In case the primary node is down, you can manually edit the $ARTIFACTORY_HOME/etc/ha-node.properties file on one of the other nodes to make it the primary node, this change requires a restart though.
3. Can you please clarify what you mean by HA separately implemented for DB and the NFS storage?
This post has NOT been accepted by the mailing list yet.
Hello, I have a similar question: Is a NFS storage mandatory for HA setup? We plan to store the binaries in a clustered MSSQL database and thus would like to avoid the need of a shared NFS storage. Is such a setup supported?
I think Ramaprakash's question 3 (HA separately implemented for DB and the NFS storage) was referring to the architecture of the database and the NFS storage: If they have to be setup in an HA manner. The answer - as far as I understand - would be yes, because both components (the database and the shared NFS storage) are mandatory for every Artifactory "cluster node" server to work properly. Thus both components have to be provided by cluster systems.