I’m thinking to set up some resiliency for our system and one solution can be having 2 clusters as main and backup which they can sync together. One option for syncing the clusters is to have couple of syncing agents which allows information to be sync in clusters. But the question is if I can add same agent to 2 different clusters or not.
There’s also possibility to create driver to do this job. Although I’m worried about CPU usage on the servers. Can you suggest any other way to sync 2 DM clusters?
Hi,
A DMA can only be part of one cluster.
If you want an element to be visible on 2 cluster then element replication can be used. More info about replicating an element can be found here
If you want the DMA to act as a backup of the other DMA then consider using failover. More info about failover can be found here
I’m not quite following what is trying to be achieved here. If a DMA needs to take over then there is the failover possibility. If you want to see the element in another cluster then there is replication. Is it the intention that if an element gets created on one cluster that it automatically appears in the other cluster and vice versa? And both clusters are active at the same time and a client can make a connection to it? If this assumption is correct then this is not possible at this moment, a failover setup is actually doing this to keep sync between both DMAs so one will take over from the other when needed. About polling elements on two clusters: when using element replication and the DMA that polls the element goes down, then the element on the second DMA also won’t be receiving new values, so not a real backup that could take over. If both clusters perform the polling then this causes more load on the device and when a parameter set is executed by one DMA it will take up to the next poll cycle before the other DMA detects the changes.
Thanks Laurens. Yes the intention is that an element gets created on one cluster and it automatically appears in the other cluster and vice versa. And both clusters are active at the same time and a client can make a connection to it.
Thank you for clarifying the situation. That will indeed not be possible at this moment as it would need to synchronise everything between both active clusters (protocols, templates, elements, views, users, settings,…), these are things that are synchronised automatically in a failover setup between a main and backup DMA, but not between two active clusters. There’s also the problem what to do with the execution of scheduled tasks or correlation rules. If both try to control the same device it could potentially result in conflicts, e.g. one DMA tries to switch to a backup device and the other DMA detects “my device is going down or is having invalid settings, I need to restore this”.
Thank you Laurens. We already have failover on the cluster. I’m looking to set up second line of redundancy in which we can have main and backup clusters. So it’s not about replicating an element. It’s more about replicating the entire cluster into another cluster.